We will discuss how hypertext is similar to, and how it differs from, other forms of information creation, organization, storage, and retrieval. We will examine the influence hypertext has on the information it organizes and the implications this has for both users and authors of hypertext systems. Finally, we want to explore a number of problematic issues, focusing on the potential for bias and distortion within hypertexts and on the paradoxical demand that hypertext systems balance accessibility with flexibility. Hypertext is already a feature of many CD-ROM's and is the basic form of the World Wide Web. These and other educational uses of hypertext will certainly be on the increase during coming years; yet the growth of hypertext has not always been accompanied by a critical reflection on its assumptions about cognition and learning, about its possible consequences for attaining certain educational benefits at the expense of other aims, or for issues of equity. We hope to initiate such a critical conversation here.
As a way of introducing readers to this new and conceptually complex learning environment, we will try to draw from various analogies and literary parallels, seeking to develop different points of connection with the basic idea of hypertext. Because the first drafts of this essay were written interactively, over electronic mail, we have also tried to maintain some of the flavor of two "voices" through the use of regular and italicized text in different sections of this essay.3 These sections can be read in a different order from that presented here, and while many of the same basic ideas will come across in any event, readers who might consider these sections in an alternative sequence would no doubt see some new connections that we did not anticipate or highlight in our original organization. We often return to the same topic or theme from different vantage points, suggesting a connection with previous and upcoming mentions of that topic. In all of this, we are trying to incorporate some elements of the hypertext model in this essay.4
What is Hypertext?
Hypertext describes a kind of informational environment in which ideas are linked to one another in multiple ways. It is a system for organizing information, just as a library card catalogue or a rolodex file are systems for organizing information:
Hypertext is a means of allowing widely differing material to coexist in a computer system; access is controlled by creating networks, links, and branches, recognizing the spatial multidimensionality of written materials, their manifold interconnectedness.5
But hypertext is more than just a new way of organizing existing information; it influences the kinds of information it organizes. As the organizing system of a hypertext grows and evolves, the structure of the information itself changes. Form and content are interdependent. This raises, therefore, deeper questions about knowledge: because knowing depends upon the meaningful organization of information, new methods of organization imply changing forms of knowledge. Furthermore, to the extent that hypertext systems incorporate the capacity both to impose patterns of organization on existing information and to facilitate the hypertext user's ability to imagine and create new patterns of organization, hypertext challenges traditional distinctions between accessing and producing new knowledge.
In hypertexts, as in texts generally, there is an interactive relation between the structure of a text and the strategies of reading it invites. While any book can be read hypertextually (a point we will return to later), hypertexts facilitate reading strategies in which the reader is making connections laterally, beyond the text, as well as linearly, within it. Hence, the discussion of hypertext also raises an opportunity to reconsider basic questions about the nature of reading texts of any sort.
History and Background
The visionary forefathers of hypertext were Vannevar Bush and Theodor Nelson. In 1945, Bush, Franklin Roosevelt's science advisor, described his idea of a memex - "an interactive encyclopedia or library" - a dynamic as opposed to static system for organizing information.6 Bush argued that this would mimic the mind's way of thinking:
The human mind...operates by association. Man [sic] cannot hope to fully duplicate this mental process artificially, but he certainly ought to be able to learn from it. One cannot hope to equal the speed and flexibility with which the mind follows an associative trail, but it should be possible to beat the mind decisively in regard to the permanence and clarity of the items resurrected from storage.7
This sort of encyclopedia or library is not only a collection of books or volumes, but a kind of "meta-book" itself - a "book of books," interconnected and cross-referenced, which can be entered from multiple points and read in various sequences. Soon after Bush, Ted Nelson coined the term "hypertext" to describe this nonlinear form of writing: "text that branches and allows choices to the reader."8 What is so novel and interesting about this type of text?
When we write text, the basic line of development is usually sequential - connecting what we said before to what we plan to say next. Our logical language and forms of organizing text ("therefore," "on the one hand, on the other hand," "Chapter 1, 2, 3, 4," "Introduction, Summary, and Conclusion," etc.) are based on the idea of a linear narrative line. Hypertext, on the other hand, is more like the novel "Hopscotch" by Julio Cortazar,9 in which the 155 short chapters can be read in different orders; the order in which they were written, or the order in which they are printed on bound pages, need not determine how they can or should be read.10 Hypertext is the next step in this development, where sentences (or chapters) do not even have to be read from the same "book," where there is no preordained order defined by the mere physical proximity of lines on a page, or by page or chapter numbers enumerating a given packet of paper. Hypertext invites "hyperreading."
Traditional Text vs. Hypertext
Traditional text assumes a kind of reading that is primarily linear and sequential. It is most often printed and permanent, produced in documents that exist as discrete physical units. Readers encounter information as it is structured by the sequence, style, and organization conceived by the author. References may be made to other texts within the narrative, but these materials exist outside the body of the document, the primary text.
A hypertext, on the other hand, is a collection of "nodes" - pieces of text that can be as small as a word or fragment, or as large as a book or another complete work - connected by "links" that allow the reader, usually in a computer environment, to access those other information sources directly from a given textual starting point. A hypertext may include a central narrative or discussion that one can read sequentially, but it also offers specific opportunities to branch off from that primary text into other textual materials. Footnotes are a simple example of this system: the small numerals at the end of certain sentences direct interested readers to explanations, citations, or elaborations available at the bottom of each page. We can extend this idea to consider, first, that these footnotes might include excerpts from other texts, or to complete texts themselves; to other sections of the primary text that one might jump ahead or back to instantaneously; to extended commentary that might be tangential to the subject of the primary text; to the marginalia of previous readers commenting on the text as they have read it; and so on. We can imagine, further, that these many supplementary textual sources are themselves interlinked through a complex system of cross-referencing. The text can be read as a hypertext. In fact, hypertext is sometimes superficially described as "electronic footnotes"; however, this description fails to capture the capacity of hypertext to incorporate richer and far more multidimensional linkages than any footnote can (including, potentially, multimedia resources).
What is the "text" that I am reading? Should the "primary text" be the main focus of concern, and the other sources merely amplifications and supplements to it; or is the primary text merely the top layer in one version, the entry point to an entire system of textual information in which, once I enter it, as in a maze, I might be led to points of discovery far afield from the primary text? Might I never return to that text, or feel the need to?
For example, if I am drawn from a chance comment in the introduction of one text to a quotation, and from that to a biography of the speaker of the quote, and from that to an historical account of the era in which she spoke, I will be not only gaining new information, but relating these "nodes" together in ways that may be entirely independent from the purposes of the original essay. As Nelson emphasizes, in this type of environment there is much greater freedom in making determinations as a reader of a text about what relates to what, or what ideas should follow or precede others: these are no longer circumscribed by the actual form or sequence of text presentation.
Indeed, hypertext begins to blur the idea of what constitutes the "primary text."11 In hypertext, the primary text exists in a web of textual information that includes annotations, references, critiques, and supplementary materials that are all immediately accessible to the reader. The interactive character of reading any text is highlighted with a hypertext, in which the primary text is a gateway into a much larger, complex network of referential material to be explored. In a hypertext, any information point should be seen, not as simply an isolated "fact" or a discrete reference point, but as a node of multiple intersecting lines of information. If we rethink this idea relationally, the year 1492, say, no longer reads as merely "Columbus comes to America" - it also reads "the expulsion of Jews from Spain"; "the completion by Michaelangelo of one of his first sculptures, 'The Battle of the Centaurs'"; and, of course, many other events, from the momentous to the trivial. This is exhilarating in one sense, mind-boggling in another. What makes Columbus in America more "central" to the importance of 1492 than Jews in Spain or Michaelangelo and his Centaurs? To whom is it "more central"? By what criteria? Did Columbus "discover" America, or "invade" it? Why 1492, not 1493? Why these Western events, and not those occurring in Africa, Asia, or the Middle East? How are these decisions made?
Analogies for Hypertext
Some familiar examples of hypertext-like systems that we use every day may help to illustrate further the ways in which hypertexts link information.12 Many of us have used note cards to record and organize bits of information. Each card contains a discrete chunk of information; these cards might refer to one another, and the order of the cards might be rearranged to reflect different views of a topic. For example, one can arrange a stack of cards chronologically or alphabetically by author, depending on one's purpose.
Reference books, such as dictionaries and encyclopedias, are also similar to hypertext. In the body of each entry, references are made to other sections of the text where additional, related information may be found. Hence the text is not simply a collection or list of entries, but also a system of interrelationships; and in fact its value as a knowledge tool may be primarily in how and where it proposes these links.
Marginalia often represent readers' attempts to establish their own links to other sources based on the associations they make with what they are reading. Comments and notes in the margins may include references to the reader's own views on the subject; to other pages or sections of the text to which the reader sees a connection; to other texts; and so on. In this process, the reader is actively modifying the text, customizing it, making it into her hypertext.
What this discussion begins to show is that in a sense every text is a rudimentary hypertext, implying choices - sometimes by the author, sometimes by the reader - about relevance, selectivity, and meaningful forms of linkage with other textual sources. Hypertexts produced with this potential explicitly in mind merely highlight an interlinked structure that is inherent to some extent in any text; the difference is that hypertexts actively invite and facilitate multiple, alternative readings of the same material.
Imagine a library with thousands and thousands of books arranged on long shelves. We take a book (it doesn't matter where we start) and connect lengths of string to every other book to which it refers. Then we take a second book and do the same, and so on. It doesn't take long to realize that this task is physically impossible. A card catalogue, including a record of books in the library, their location, and a large number of cross-references, is a fairly complex hypertext, but even its information storage capacities are limited by space (how large each card is, how much information it can contain) and accessibility (how many cards there are, how long it would take to leaf through them to find the reference one seeks). But an on-line computer hypertext makes possible the virtually limitless and very rapid linking of information.13
Electronic links are not subject to the same physical or practical limits of proximity (a major issue in libraries is what you place near other things), so the links can be multiplied enormously.14 But there is a further level of complexity. In this first-level example, we were only connecting texts to other texts to which they actually refer. But then, in addition to these connections, there are new connections we might discover in the process of exploring the texts. Each link generates many potential new links. So nodes A and B are linked. B was linked to C, so now A is linked to C. But the link of A to C also changes the link of C to B; or gives rise to a new node ("C in light of A"). Making linkages within hypertext is not simply a matter of mechanically associating givens; it is inevitably an active process of interpretation - of authoring - itself. And this process of interpretation is inevitably open-ended and indeterminate, because of the "hermeneutic circle" inherent in forming such linkages: the cycle of how novel information changes the way we understand familiar information, thereby making it novel information, and so on. Because this process is so unpredictable and potentially infinite, it even begins to blur the very notion of significant, discrete data points: what do A, B, or C mean when they come to be seen in immediate relation to one another - and, in turn, to many other things?
Gilles Deleuze and Fˇlix Guattari suggest a powerful metaphor for this construction of the text in their discussion of rhizomes.15 A rhizomatic plant (mint, certain grasses, etc.) depends upon a root-like system that is decentered, spreading in all directions; a more tree-like plant structure, on the other hand, depends upon a central tap root and a more hierarchically organized set of roots of progressing size, centrality, and importance to the system ("any point on a rhizome can be connected with any other, and must be. This is very different from a tree or root, which fixes a point and thus an order"16). Deleuze and Guattari develop these two metaphors, the rhizome, or grass, versus the root system, or tree, into two guiding models of the text. They describe the rhizomatic in terms of six principles:
1 and 2. Principles of connection and heterogeneity: any point of a rhizome can be connected to any other, and must be....A rhizome ceaselessly establishes connections between semiotic chains, organizations of power, and circumstances relative to the arts, sciences, and social struggles....
3. Principle of multiplicity: it is only when the multiple is effectively treated as a substantive, "multiplicity," that it ceases to have any relation to the One....Multiplicities are rhizomatic....A multiplicity has nether subject nor object, only determinations, magnitudes, and dimensions that cannot increase in number without multiplicity changing in nature....
4. Principle of asignifying rupture: against the oversignifying breaks separating structures or cutting across a single structure. A rhizome may be broken, shattered at a single spot, but it will start up again on one of its old lines, or on new lines....These lines always tie back to one another. This is why one can never posit a dualism or a dichotomy, even in the rudimentary form of the good and the bad....
5 and 6. Principle of cartography and decalcomania: a rhizome is not amenable to any structural or generative model. It is a stranger to any idea of genetic axis or deep structure....A rhizome is altogether different, a map and not a tracing.17
This discussion illuminates a key feature of poststructural thinking, but its main value in the present context is to suggest a useful way of thinking about hypertext - as a rhizomatic textual system. 18While we began with the idea of a root, or "primary text," from which all other references were merely extensions and addenda, we found that very quickly the multidimensional structure of hypertext puts all references and texts at a common level, no one of which can claim a priori centrality, but only a relative centrality, given a particular purpose. Even hierarchically organized hypertexts can be given over to very different uses; it is the nature of any hypertext system to permit exploration across unanticipated pathways that belie any pregiven organization or hierarchy. Hence this rhizomatic structure can be seen as both a feature of the organization of text, and as a way of reading any text nonlinearly and nonhierarchically - the difference is the degree to which a hypertext, by explicitly representing such a nonlinear, nonhierarchical structure, encourages such readings.
The significance of computer hypertext systems, as opposed to other, more simple hypertexts, lies in the number of associations that can be made and the ease and speed with which they can be accessed. Only by means of a computer network can one have direct access to every other linked node in a system.19 When one is reading a book and it refers to a passage in another book one may, depending on one's degree of interest, get up, walk to the library, find the book, check it out, return to the office, find the passage, read it, and then (perhaps) return to the original text (maybe the second book turns out to be more interesting or useful). But, of course, that book might not be available, or it might be raining. In a hypertext system, the same reference can appear on the computer screen at the stroke of a key - as easily and automatically as one can call up the following page or any other part of the original text. As discussed before, this reference may lead to a third reference, or back to the original material. This seamless shifting from text to text is only possible on-line. The sequence of page or chapter order within a text becomes only one of many possible organizing systems.
Moreover, hypertext throws open the parameters of what can be searched for. In traditional forms of data organization, the search parameters are fixed. In a library's card catalogue, for example, access to text and information are restricted to certain fixed search categories such as author, title, or previously codified descriptors and keywords. Hypertext is able to operate among any segments of text, allowing them to be accessed in flexible, intuitive ways.20 Moreover, in at least certain types of hypertext, these links are not only passive (hard-wired into the system) but active, allowing readers to create new links, and new types of links, in terms of their own emergent understandings of the material. The immediacy and flexibility of hypertext are, therefore, much more than matters of convenience. At a deeper level, hypertext can not only allow accessing author-identified references, but also allow readers to create and search their own associative references. In this, they become not only consumers of a text, but active contributors to it: the distinction of author and reader, as Michel Foucault and many others have pointed out, begins to break down.21
Hypertext, Knowledge, and Thought
As Vannevar Bush noted, the structure of hypertext environments parallels and can facilitate the ways in which we learn: nonsequentially, dynamically, and interactively, through associations and by exploration. Hypertext can allow the user the freedom to navigate courses through the material in a manner determined by his or her own interest, curiosity, and experience, or by the nature of the task at hand, rather than following a course predetermined by the author. Hypertext makes concrete the idea of interactive reading.
This process of actively selecting and assimilating new information in light of personally coherent cognitive frameworks meshes the potential of hypertext with constructivist learning theories, especially schema theory. This link is particularly strong when we consider knowledge domains that are complex and indeterminate; domains requiring a high degree of "cognitive flexibility" and a tolerance for ambiguity.22
"Learning" and "understanding" operate by making connections. We come to comprehend something when we can bring it into association with other things we already know. Mind and memory are themselves hyperenvironments: we do not learn new information as discrete, isolated facts - or, if we do, we are not likely to remember them for long. The information we learn best is material that can be integrated with knowledge we already have, frequently through complex and multiple links of association. John Slatin captures this idea nicely in discussing the need for writing that "surprises" the reader:
The informational value of a given document is not simply a function of the quantity of information it presents or the facts it contains. At one level of abstraction, what we call information may indeed consist in numbers, dates, and other information, other facts....At a somewhat higher level of abstraction...none of these data can be considered information until they have been contextualized, arranged in such a way that both the significant differences and the significant relationships among them may become apparent to the intended reader....[T]his is when information becomes knowledge.23
In this process of establishing active, novel, and idiosyncratic patterns of association, the line between "primary" and supplemental materials fades and disappears. Even the conventional bonds that hold together what we once would have considered a "single" text begin to dissolve. As Paul Delaney and George Landow describe it,
the text [in a hyperdocument] appears to break down, to fragment and atomize into constituent elements (the lexia or blocks of text), and these reading units take on a life of their own as they become more self-contained and less dependent on what comes before or after in a linear succession.24
For example, there are lines from literature and elsewhere - "'Twas brillig," or "Fourscore and seven years ago" - even scenes from movies - Cary Grant hanging off Mt. Rushmore, Humphrey Bogart saying good-bye in Casablanca, etc.,- that evoke an entire range of responses, images, and associations, quite apart from the narrative in which they were originally contained. They stand between that text and the nodes of the association they evoke. In that sense, the "primary work" has broken down into a series of independent and separable nodes.
In this view of hypertextual representations of knowledge, all nodes become "leveled" - none are a priori more important, or central, than any others. This is the poststructuralist dream state: a limitless bricolage of fragments and pieces that can be brought into new and unpredictable associations with one another.25 The possibilities of novelty and creativity inherent in this view of knowledge should be clear, as is its potential for chaos, arbitrariness, and spinning out endless permutations and juxtapositions for their own sake. 26The development of computer-based hypertext systems is not responsible for this shift in our understanding of the text and processes of reading - the work of Roland Barthes, Umberto Eco, Michel Foucault, Jacques Derrida, Hˇl¸ne Cixous, and others anticipate these technologies of reading and authorship27 - but hypertext instantiates this decentered view of the text and a corresponding shift in authority from author to reader in the interpretive process: "Hypertext has no center...[which] means that anyone who uses hypertext makes his or her own interests the de facto organizing principle (or center) for the investigation at the moment. One experiences hypertext as an infinitely decenterable and recenterable system."28
But now we begin to see a paradox (the first of several in this discussion): If hypertexts do make possible the manifold linkage of nodes to different points of association, they also have the effect of fragmenting and decontextualizing each node, freed up from its position in some original narrative or line of argument. "Lateral" associations may turn out to be more useful in certain contexts than the original "linear" ones; but the leveling of all associations without "privileging" any particular one may make every association appear arbitrary. Yet it does not seem arbitrary, for example, to "privilege" the idea of an historical sequence, or of a literary story line, or of a logical explanation, as particular ways of relating information "nodes" (even if one might acknowledge that alternative ways of organizing the same information are also possible). Without some such starting point, even a provisional one, the exploration of a rhizomatic system might be simply anarchic - a nice image for avant garde literary interpretations, perhaps, but not necessarily for beginning learners. Problematizing given narratives is one thing; proposing alternative readings is another; but the very constructivist theory that encourages these also explains why a reading with no center is no reading at all.
Writing and Reading Hypertext
Within a hypertext environment traditional distinctions between authoring and reading break down still further. Authoring involves making a variety of decisions about where a text begins, where it ends, and the order of textual elements from start to finish (although some authors, such as Cortazar, have playfully abandoned some of that authority). Authors must be particularly concerned with the arranging and sequencing of information or arguments - ordering the conceptual links in a linear sequence. Authoring also involves decisions about what materials become part of the main document and what are relegated to references, footnotes, textual allusions, and so on (or what is left out of the discussion entirely).
The production of traditional text tends to be exclusive as opposed to inclusive. We spend much of our time writing, at one level or another, deciding what to leave out. Given finite space limitations, and often finite time, we need to apply fairly strict self-discipline in authoring a text: because this text can only address certain topics, and not the many others that might be interesting, relevant, and important, but for which there is neither time nor room. Hypertexts remove some of these limitations: virtually anything that is judged interesting, relevant, and important - from any standpoint - can be included and made accessible to the reader. In hypertext, the premium can be on drawing more and more sources in, multiplying the number of data points and diversifying the direction of meaningful associations - to a potentially limitless (and perhaps at some point counterproductive) degree.29 Umberto Eco calls this a process of "unlimited semiosis."30
Authors of hypertexts will need to produce their work with an eye toward how it will fit within a transformed system of reading. Authors, of course, may still write out sentences or compose pages of prose. But in the final text, traditional concerns for beginnings, endings, order, and sequencing are vastly complicated, as concerns of navigability through entrance and exit points that will be useful to prospective readers are multiplied.31
Stanley Fish tells a well-known story, very apt in the present context, in his book, Is There a Text in this Class? He recounts teaching a class in English religious poetry of the seventeenth century. As students enter the room, Fish notices a vertical list of names/words on the blackboard, left over from the previous class: "Jacobs-Rosenbaum, Levin, Thorne, Hayes, Ohman." He adds the descriptor "p. 43," and tells the English lit class that this is a religious poem and asks them to interpret it, which they proceed to do with gusto. The lesson he draws from this example, of course, is that "Interpretation is not the art of construing but the art of constructing. Interpreters do not decode poems; they make them."32 Virtually any finite set of nodes, it appears, can be subject to some sort of meaningful interpretation. Who is the "author" of the poem on the blackboard: the previous instructor? Fish? the students in the English lit class? Is it a "poem"?
The demarcation of authors and readers in hypertext environments is similarly problematic. On the one hand, the author's capacity to impose unilaterally a necessary structure and sequence on a text is undermined as the network of links becomes more and more complex; conversely, the process of reading involves the active making of linkages between nodes of text in the same way that "authoring" does. The relation of author and reader is made reciprocal: the "accessing" of textual information influences its "production," and not only vice versa. Readers become authors, and authors need to view their own productions as readers.
Furthermore, the demarcation of primary texts and commentaries or interpretations on those texts also becomes unclear. The author's capacity to favor a specific interpretation or meaning to a text is undermined when the process of reading involves the active making of linkages between nodes of text in the same way that "authoring" does. Novel interpretations, in a hypertext environment, can actually be manifested as alternative ways of organizing and representing textual material: the reader can literally, and fairly easily, impose an entirely novel and idiosyncratic order upon the original textual materials. The interpretative process becomes not only a lens, but also a filter, sifting and shaping what one sees as relevant. A hypertext system, therefore, develops an ambiguous relation to any primary texts that might be part of it - incorporating them, commenting upon them, but also altering them in the process.
A brilliant example of this process, and a frequently cited precursor to hypertext, is Roland Barthes' book, S/Z, an exhaustively detailed analysis and commentary on a short novella by Balzac, "Sarrasine."33 Although the original novella is only a few dozen pages, Barthes breaks it up into more than 500 separate text units, or lexias, each of which he discusses at length, then cross-references them in an astonishingly intricate manner, producing in the process a parallel text that dwarfs the original. Barthes' book, in fact, reads fairly well on its own, without knowing anything about Balzac or his novella; and of course the novella can be read without the assistance of Barthes. Which is the primary text here? Which serves to amplify and illustrate the other? Having read "Sarrasine" through the readerly/writerly interpretive network of Barthes, can one ever simply read the novella on one's own?
There is a strange passage in Small World, a novel by David Lodge, about a literary critic studying "the influence of T.S. Eliot on Shakespeare."34 In a hypertext environment, this can make sense, because the actual historical sequence of producing texts is not the only constraint on reading. The point is not that Eliot (the living person) could have influenced Shakespeare (the living person), of course, but that our understanding of Eliot influences our understanding of Shakespeare. We cannot read Shakespeare today without echoes of Eliot coloring our readings (compare, in this sense, the "influence of Barthes upon Balzac"). A similar point is made by that most hypertextual of writers, Jorge Luis Borges; in his short essay "Kafka and His Precursors," he suggests that rereading previous authors with a contemporary understanding of their influence on Kafka in mind creates not only a new reading of Kafka, but a new reading of these texts in relation to one another:
If I am not mistaken, the heterogeneous pieces I have enumerated resemble Kafka; if I am not mistaken, not all of them resemble each other. This second fact is the more significant. In each of these texts we find Kafka's idiosyncrasy to a greater or lesser degree, but if Kafka had never written a line, we would not perceive this quality; in other words it would not exist....The fact is that every writer creates his own precursors.35
A hypertextual history of literature enables us to incorporate and to express just such complex lines of influence and cross-reading.
It is no longer even clear that it is useful to think any longer strictly in terms of single texts. Today, it still makes sense to talk in terms of "reading a book" - Simone de Beauvoir's The Second Sex, for example. One starts on page one and continues through to the end. But in a new hypertext edition, this volume might be reproduced along with the original French (Le Deuxi¸me Sexe), so that the reader can move effortlessly back and forth between the French text and the English translation. Alternatively, de Beauvoir's book might be produced alongside Betty Friedan's The Feminine Mystique in an edition that cross-references passages from both books, emphasizing points of similarity and contrast between them. One might read such a hypertext, not linearly within each book, but in a criss-cross of moving back and forth between the original texts, forward and back with each text, and only indirectly getting a sense of the narrative line within each book, separately.36 What book are we reading: de Beauvoir? Friedan? or something new? This sort of reading implies a very different orientation to any given text or author, and raises a host of new questions about what it means to "read" effectively.
What these examples show even more clearly is that "hypertext" is actually a hybrid term, meaning both the particular technological developments that have made textual fragmentation and complex cross-referencing possible and convenient; and a theoretical view of the text (any text) as decentered and open-ended, which has been with us for a long time. What we have called "hyperreading" is both the kind of reading that hypertexts tend to encourage, and a more general view of reading as active and deconstructive/reconstructive, which can brought to bear, to some extent, on any sort of text. Even though authors will continue to produce things that look like traditional texts, they will increasingly have to come to grips with the consequences of how these texts will be read and interpreted by a hypertextually-aware readership.
To the extent that one is purposely constructing a hypertext, one is performing an operation upon the textual elements of which it is composed (some of which one might have written one's self). As should be quite clear by now, this is an active, interpretive process: the organizers of textual information are knowledge producers - they are authors of an important sort. Librarians, archivists, cataloguers, indexers, editors, translators, anthologists, and so on, are not merely in the process of archiving, organizing, or providing access to information, but are themselves producing texts: hypertexts. They create systems that associate new textual information to other textual information. None of these activities is new, of course, but their role has become even more crucial as the volume of information available grows exponentially and as the technical means of accessing and organizing information become more powerful and complex.
These activities, in a technologically driven information system, move from being facilitative (which they always were) to being indispensable. In all fields of inquiry, scholarship, literary production, and commentary, the sheer volume of textual material available has exploded; and with this comes an accelerating rapidity with which it is produced, consumed, and becomes obsolete (a problem most clear in certain scientific and technical fields, where by the time a new research article reaches print its information is often already outdated).37 No one can read everything relevant; and not everything relevant might be worth reading. With the enormous growth in volume of textual materials comes a greater need for selectivity; how does one decide what is relevant and what is most worth reading? These changes mean that the selection, evaluation, and organization of new information, in a form that is accessible and useful to readers, has become increasingly the responsibility of people who are intermediary between author and reader. These textual intermediaries (as we saw with Barthes, for example) must be both educated readers and authors themselves, producing hypertexts that compile, relate, and interweave the elements of different texts in a meaningful way. And, as we will discuss below, once we recognize this shift in responsibilities, the question of who these intermediaries are, and how they make these judgments, comes to be seen as a crucial intellectual and social problem, especially for education. One important dimension of this question is whether such authors regard themselves as teachers (or, conversely, whether teachers will play a role in such authoring): that is, whether hypertexts will be created with an understanding of the cognitive and developmental needs of diverse readers (students) in mind, or whether they will be structured in terms of the associations of knowledgeable experts, regardless of whether these will be accessible to a nonexpert audience. A related issue, which we have mentioned and to which we will return, is whether these links are "passive," presented as givens within the hypertext, or "active," allowing readers to create their own new links as they come to understand the material better and in distinctive ways. With libraries, decisions about the organization of texts were made by one group on behalf of others, and had to be made in one consistent filing and accessing system, with all of the implicit assumptions or gaps that system might have entailed. Hypertexts allow for multiple, and partly user-generated, forms of organization - although it remains to be seen whether this will be the predominant form they will take.
In such an environment, the new inventions of knowledge will be heuristics: useful ways of putting things together in the face of a morass of overwhelming information. So "chronological sequence" might be one kind of heuristic; "causal relations" another; "analogical similarity" another, and so on. These are the kinds of interpretive tools that the typical reader of hypertext will need in order to find and access meaningful information related to the text she happens to be reading. Moreover, the question of how the "interface" is designed for a hypertext will influence its usefulness and accessibility; many readers will need to be able to call up for reference explanatory materials that make explicit to them the implicit structures of the hypertext system. Indexes, charts, maps, glossaries, concordances, search engines, and so on, become more than just guides to moving around a hypertext; they become crucial textual elements themselves, replete with their own interpretive assumptions, emphases, and omissions.
A number of writers have addressed these and similar concerns about the need for a "hypertext grammar."38 Delaney and Landow, for example, identify the need for "stylistic and rhetorical devices" that orient readers to where they are in the hyperdocument, help them to read and navigate efficiently, indicate where links lead, and assist readers who have just entered the document to feel "at home" there.39 One of the best examples in some current hypertext systems is familiarly called a "bookmark": a means for users to tag some important element (important to them, that is) in the hyperenvironment so that they can return to it directly, rather than by remembering and retracing the exact pathway through which they came across it. This is one simple way in which users can customize a hypertextual environment. David Jonassen also raises issues of navigation: How does one get about in the document? Where are its entrance and exit points? How does the point at which one enters the hypertext influence the user's subsequent understanding of the material? How structured must hypertexts be? What risks are there that the user will suffer from "cognitive overload," given the potential richness of the hyperenvironment? How can authors anticipate and alleviate this result?40
We believe that addressing such questions of organization in hypertext will depend upon distinctions between the different sorts of readers who will encounter hypertexts, in educational and other settings: following Slatin, we will call them browsers, users, and co-authors.41 Although there are overlapping cases, and the difference is mainly one of degree, these terms describe three basically different readerly orientations toward hypertext systems, especially in the readers' need for and use of explicit guidance toward specific associations among textual elements and in their capacities to identify and establish novel associations among textual elements in their own active processes of reading. Furthermore, people who are users and co-authors in certain environments, dealing with material about which they have more background knowledge, may be browsers in others.
Browsers, for example, represent casual, curious readers. The signs or navigational aids that are available may mean little to them because browsers are doing just that, browsing. As Slatin points out, because the pathways of certain hypertexts are nested hierarchically, it is probably less important to anticipate where these types of readers will go than to provide them a means to backtrack when they get lost.42 Hence an especially important feature in their textual universe will be a list or map of the previous selections they have made, in the order in which they made them; this list or map can be called up by the reader if she wants to return to some previous point in the journey. Browsers may look at many text elements, but they are not actively seeking to create associations or patterns among them, nor do they need to know how to make any changes or additions to what they find.
Users, on the other hand, have reasonably clear ideas about what they are looking for. Often seeking some specific information from the hypertext, these users need quite precise directional information - signs to indicate where certain branching options might lead and what the user will find there. This raises authoring and organizational questions about hypertexts and how linking paths can be created that give adequate directions, useful heuristics, and a necessary degree of predictability, while also being flexible enough that more experienced and knowledgeable readers can move freely between different sources.
Co-authors place much greater demands on the hypertext, since they require not only the resources and guides to move about within the system, but also the means to actively - and perhaps permanently - change and add to the system in light of their own active reading. As discussed previously, many academic readers do this to some extent now. We highlight or underline text, write notes in the margins, keep annotated bibliographies, or copy and file relevant quotes and passages from the works we read. In a hypertext, all of these alterations and additions can become part of the hyperdocument - if the environment is structured "dialogically," in the sense that it allows for an active response and intervention by the reader (as opposed to a "read only" mode that does not allow modification or custom tailoring). In our view, it is important that hypertexts not only permit authors to establish the context in which information becomes meaningful, but that they also allow active, knowledgeable readers to construct and record their own meaningful links - and in this, to become co-authors.
Hence, designers make a fundamental choice when creating hypertexts. We have distinguished between texts that are more static, or passive, and those that are more dialogical, or interactive.43 In the former, the links and pathways are set; they cannot be changed, nor can new ones be constructed by the readers. The built-in links of access to other materials may be very complex, and they may be explorable along myriad alternative routes (such as within the World Wide Web), but they are restricted to those anticipated and constructed by the designers. Clearly, this form of hypertext makes the most sense for relatively inexperienced or less knowledgeable readers, those we have called "browsers" and "users."
In this context, the nature of those links that are "hard-wired" into the hypertext becomes crucial. Some authors argue that these should anticipate the most natural and intuitive links for likely users: "The authoring challenge is to design the structure of the hypertext database to match the ways that a user might want to think about the topics....Knowledge must be structured in a way that supports the mental models that readers may create when they use the hypertext system."44
Others argue, on the contrary, that the structures of hypertext should mimic "expert" understandings of the material or an inherent structure of the discipline: "The organizational structure of hypertext may reflect the organizational structure of the subject matter or the semantic network of an expert....If we assume that learning is the process of replicating the expert's knowledge structure in the learner's knowledge structure, then learning should be facilitated by a hypertext that replicates the expert's knowledge."45 Some options within computerized hypertexts not only involve linking specific nodes, but also performing operations on them: searching, sorting, filtering, prioritizing, and so on. Often these options too are not discretionary; they represent the organizational heuristics that designers believe are most useful, which may or not conform to those most beneficial or meaningful to readers.
It is not within the scope of this project to adjudicate this debate; and as some have pointed out, certain authors seem to advocate both these views simultaneously.46 As readers already will have seen, each of these approaches is subject to some dangerous abuses. But we do mean to point out the fundamental tension between them, and the fact that they manifest two crucially different conceptions of learning: once one is committed to a "passive" system in hypertext, this choice reflects a bias toward particular learning possibilities inherent to the type of system being developed.
What we are calling "dialogical" hypertexts, on the other hand, are characterized by more interactive and flexible constructions. Their main virtue is a capacity for rhizomatic growth. In a highly interactive hypertext, we lose the traditional notion of a single authorial perspective altogether. We have suggested the value of hypertexts in which reader contributions, such as annotations, marginalia, evaluations, critiques, new textual nodes, and new linkages, all can become temporarily or permanently part of a customized system. The hypertext becomes a dynamic, organic growing body of information; an evolving representation of, and spark to, the reader's own developing knowledge and understanding.47
In this dialogical view of hypertext, the inherent openness of the text can be turned to advantage. If it is generally true of reading that a reader imposes an idiosyncratic texture or order on what she takes from a text; if it is generally true that a hermeneutic cycle of interpretation continually changes the understanding of each part as the reader comes to understand others; and if it is generally true that every text is implicitly linked (or linkable) with a virtually infinite variety of other texts; then the technical capacities of hypertext, which as we have argued are especially suited to incorporate and organize such textual openness, should be exploited to structure it intelligibly. But for this to happen, there needs to be some procedure for readers to alter what the hypertext author provides them.48
A key question, in this regard, is whether these changes should be recorded only in a personal, customized version of the hypertext (for example, on a CD-ROM or other storage system on the user's own computer); or whether these should be compiled permanently as part of a master system - accessed by multiple users as part of an on-line network, for example - so that the hypertext would be a continually growing, changing system, in which new readers could access, among other things, the recorded comments, additions, and linkages of previous users. In some ways this instantiates a more democratic, social, decentralized view of text production; yet it also threatens to devolve any hypertext rapidly into a mess of unwieldy size and chaotic complexity. It seems that a useful hypertext will inevitably reflect to some extent the more or less predictable, coherent, and finite choices of a specific author, editor, translator, educator, etc. (or a group of these, working in concert) - and in reading any particular hypertext, one gains the benefits and suffers the limitations of their interpretive insights.
Hypertext and the Learner
The Design of Hypertexts
Unfortunately, the desire to structure a hypertext in an open, dialogical fashion encounters a difficulty when we look at the concrete problems of the learner, and of the different types of readers who might encounter a hypertext. A form of organization that only allows a novice to search through direct and explicit connections may not facilitate the development of that novice into an independent and autonomous reader who can alter and add to what he or she finds in a hypertext. Conversely, a dialogical and flexible hypertext system, of much use to those who are prepared to be contributing co-authors of a text, might be too open-ended to be of much use to a novice or to a user who is simply interested in extracting specific and already-organized information from the textual source. Such choices between accessibility and flexibility reflect implicit decisions about learning styles and about the audiences to whom hypertextual materials will be practically useful. These are educational decisions, but also social and moral ones.
A major concern in the literature on hypertext is with the experience of novice readers getting "lost in hyperspace," following a meandering path of associations into the hypertext field, then finding that like Hansel and Gretel they cannot retrace their steps back. In some cases, the sheer volume of information, and the number and flexibility of pathways that are available, simply become overwhelming. A substantial body of research suggests that this experience becomes a major source of confusion and frustration, often discouraging new readers from experimenting further with the system.49 While, as noted, there are features that can be added to a hypertext system that provide an exact record of the path a reader has traced to arrive at a given node, this aid in itself does not always help readers to understand how they arrived there, or where to go next. As a result, many readers of hypertext end up browsing or performing the textual equivalent of "channel surfing": quickly scanning or surveying randomly accessed information, in very short snippets, with no overall sense of coherence or meaning for what they are exposed to.
In a sense, this is an updated version of "Meno's paradox." Originally, Plato asked how learners can ever learn anything truly new, since if it is entirely unrelated to what they already know, it will not make sense to them; but if it is closely linked with what they already understand, or can deduce, then in some sense they already "know" it, and are merely recognizing it. The hypertext version of this paradox is: How do you look for something if you don't already know what it is or where it is? A novice encountering a complex hypertext system for the first time cannot possibly know what information the system contains, without happening to come across it through searching or guesswork. But if there are explicit guides built into the system that direct the novice to particular information points (as most traditional texts do, through the organization of pages, chapters, tables of contents, and the use of explicit connectives, such as "therefore," "in conclusion," and so forth), then the danger is that the novice may become dependent upon this particular system of organization, and not become capable of developing his or her own.
The educational implications of this issue are profound. Beyond allowing students to proceed through the document by taking prescribed routes, in a specific sequence, at a deliberate pace (the once-heralded attributes of computer-aided individualized instruction), hypertext can permit students to focus their investigations on questions informed by their own particular interests and experiences. They proceed through and organize materials in ways that make sense to them, developing their own heuristics. This flexibility has many advantages, not the least of which is a capacity to accommodate different personal or cultural learning styles. But in order to reach this stage, learners need to have experience with explicit tutorials, guides, indexes, and so on, that provide them both with a basic grasp of what the textual source contains and with models or heuristics that they can learn from and adapt as they become more autonomous readers of hypertext.
The danger that hypertexts may be too unstructured to accommodate the needs of learners has led some to alarm:
It's a way of presenting documents on the screen without imposing a linear start-to-finish order. Disembodied paragraphs are linked by theme; after reading one about the First World War, for example, you might be able to choose another about the technology of battleships, or the life of Woodrow Wilson, or hemlines in the '20s. This is another cute idea that is good in minor ways and terrible in major ones. Teaching children to understand the orderly unfolding of a plot or a logical argument is a crucial part of education. Authors don't merely agglomerate paragraphs; they work hard to make the narrative read a certain way, prove a particular point. To turn a book or document into hypertext is to invite readers to ignore exactly what counts - the story.50
Yet this argument shows, at most, that learning certain conventions of linear narrative and argumentation should represent an important phase of students' learning. These are not the only useful means of interpreting and organizing information, and for many purposes they can be counterproductive. Hypertexts are educationally valuable because they highlight possibilities that are inherent to the processes of reading and thinking: the possibility of constructing a unique, personally meaningful, and useful interpretation of textual materials; and the susceptibility of information to more than only one form of organization. Yet they also teach that not just any organization will do; that there are certain conventions and heuristics that promote meaningful and useful interpretations. Associating the causes of war with the movements of hemlines may appear frivolous, and it may prove to be just a random collocation; but it might also lead to a novel and revealing understanding of the links between, say, militarism and changing gender identities. It is in learning to tell the difference between heuristics or associations that help to support meaningful and useful interpretations, and those that do not, that the real work of education needs to proceed. Hypertexts that allow a degree of unstructured and idiosyncratic exploration can be an indispensable support to such learning; and hypertexts with a degree of structure built in, but also the options of customized design, may serve as effective bridges or scaffolds to bring readers to the point where they can create more personal and distinctive organizations of the textual material available.
When readers are involved with making links, and not only exploring along pathways established for them, this may also encourage the kind of "metacognitive" awareness that recognizes alternative forms of organization for information and their relative merits for different purposes; and it may develop a greater self-understanding in learners about the forms of organization that they tend to favor, and others that may be possible. It will allow hypertexts to be used for different purposes than designers could ever imagine; this is important not only because it may allow hypertexts to be read "against the grain," but because allowing texts to be read in multiple ways is one way to involve readers who might otherwise never have or desire access to those texts.
Hence, the decision between structure and freedom is not necessarily an either/or matter, and it is possible to imagine hypertext systems that may incorporate capabilities for both prestructured and personally structured readings; but doing so effectively and fairly will require remaining sensitive to questions of learning, diversity, and the nonneutral relationship between systems of organizing knowledge and patterns of affiliation or exclusion for varied audiences.
The Content of Hypertexts
Most of the discussion up to now has stressed the form of organization that hypertext uses, in order to highlight it as a system of knowledge in its own right. However, despite the deep interrelation of form and content of knowledge, it remains important also to examine what textual content is included in hypertext systems, and what is left out. Will such decisions be guided by Hirschian assumptions about "cultural literacy," and what every citizen needs to know; or will they tend toward multicultural sources, represented in all their diversity? Will they privilege certain texts (for example, The Bible) as more central textual reference points, while relegating other texts (for example, The Koran) to more peripheral status? Will they, to use the metaphor of Deleuze and Guattari, be more root-like (hierarchical) or more rhizomatic (decentered)? Hypertexts may make conceptually possible the representation of a decentered, multifaceted view of knowledge, of how we come to know, and of what we need to know; but it does not follow from this potential that the actual designers of hypertext will allow that potential full play.
Hypertext, as should be clear, can be an enormously liberating innovation or a powerful system of ideological hegemony. Too many writings on hypertext, as with so many writings on educational technology in the past, adopt an overly celebratory tone, heedless of the potential for manipulation and control built into any powerful technological system.51
As discussed previously, all texts require decisions about selectivity, relevance, and implicit or explicit structure. No text can contain everything, and the more that is included, the more crucial become choices about organization and connectivity. It appears that hypertext alleviates some of these concerns: the sheer volume of information that can be included, the variety of textual types that can be accessed, and the multiplicity of connections that can be made, all suggest that designers will not need to exercise such severe constraints. And, indeed, to an extent this is true. But the problem simply appears at a deeper, more subtle, and hence potentially more insidious level: that the very virtues of hypertext - complexity and comprehensiveness - make implicit authorial, organizational choices all the more essential for the usability of texts, and yet all the more difficult for the reader to detect. Significant absences, silences, or exclusions in hypertexts are unlikely to be seen or noticed by the vast majority of their users; and the very complexity of the heuristics designed by hypertext authors will make them all the harder to diagnose and criticize, except by readers of comparable expertise - who, by definition, will be rare.52
The Designers of Hypertexts
If no system of organization in hypertext is neutral, and if the heuristics incorporated into a hypertext always express implicit intellectual or moral priorities, then the question of who creates them, with whatever biases they might contain, becomes a critical problem. For example, Dale Spender has suggested that women should be suspicious of a technology so overwhelmingly controlled by men, because its systems of knowledge will express the inevitable ideological limitations of the male perspective on questions of sexuality, gender relations, and so on: she asks, for instance, where the controversial topic of abortion would be situated in a complex data environment, to what nodes it would be linked (with "murder," or with "human rights"?), and hence by what cross-references it might be accessed.53
Authorship, with hypertexts as well as all other sorts of text, raises questions of authority: who will constitute this new class of knowledge producers/knowledge organizers? Cataloguers, archivists, and others who construct hypertexts are not just compiling and organizing a separately existing body of knowledge; they are acting as interpreters and makers of knowledge themselves. As the autonomy and status of individual texts breaks down, the inclusion and exclusion of materials in hypertexts, and the links created between them, will become far more influential than any single component in the system - especially in educational contexts.
The kind of technology necessary for storing and manipulating the vast amounts of information used in creating hypertexts is expensive and highly sophisticated. The ability to operate and manage these hardware and software systems will begin to define a new class of knowledge producers and organizers. Who these new experts are, how they are educated and selected, and what epistemological, social, or political views they hold, will become crucial questions. It has often been claimed that access to knowledge will become more democratic with the advent of personal computers. And the explosion of data sources, discussion groups, web pages, and other textual materials available over the Internet has certainly put more information into circulation and given more people access to it and the opportunity to contribute to it. But with this very explosion comes a new need for indexes, search devices, and filters that help users find and sort through the enormous volume of information available; and this in turn depends in many cases on an intermediary, visible or invisible, deciding what will and won't be made available, and how it will be organized.
Whether we name them in terms of benign names such as "archivist," "custodian," "gatekeeper" - or "teacher" - those who create hypertexts will control information and access to it in ways that are potentially much less democratic and more restrictive and hegemonic than is now possible with simpler informational systems. Just as our political democracy is not direct but representative (relying on specialists, bureaucrats, regulators, and other elected or unelected officials who are theoretically "accountable" but who in practice operate most of the time outside public scrutiny), access to knowledge is becoming more democratic only to the extent that this informational democracy rests on the presumption of a group of intellectual authorities and experts whom we trust to make prudent, fair, reasonable, and socially useful decisions - but who in fact are rarely susceptible to public challenge and question.
Michael Koenig tells the story of a physician who searches a computer information service to see if a certain drug regime could be causing jaundice-like symptoms in one of his patients. In the end, he changes the patient's medications based solely on the title of an article. He reads neither the abstract nor the text, but bases his decision on the fact that the article appeared in the British Medical Journal.54 To what extent should we base substantive decisions about belief and proper courses of action on the reliability of choices made by others unknown to us? To what extent does a hypertext environment require us to do just this, over and over again?
The typical reader of hypertext will often find it impossible to judge independently the value or credibility of the material she discovers. In hypertexts it may be difficult to attribute a chunk of text to a particular author, because of the ways primary texts are deconstructed. One might know that Professor Jones is a leading expert on the Civil War, but in a hypertext about the Civil War may find it difficult to distinguish her contributions from anyone else's (or one might not know who Jones is or whether she is credible or not). The leveling effect of hypertext on multiple information sources, praised earlier as providing an opportunity for interpretive flexibility and open-endedness, now comes back to haunt us: in many cases we want, we need, someone to decide for us what is valuable, what is credible, what is worth our time and attention. Does an editor or reviewer make those decisions? If so, it might raise concerns that too much power is invested in certain individuals or groups. By contrast, one can record the numbers of previous users who have accessed a particular source, and perhaps also their comments, as a means of providing some criteria of reliability for future readers - but in the end, this simply shifts questions of reliability from one set of doubts to another.
A situation where the organization and dissemination of knowledge are controlled by a few is exacerbated by a highly bureaucratic and political environment in which governmental or private funding for large archiving and editorial projects comes at the cost of compromises. To what extent will those who fund the development of such sophisticated systems expect them to be congruent with their own interests? For example, will R.J.R. Nabisco be willing to provide grants for hypertexts to be used in health classes, which might be concerned, in part, with the adverse effects of smoking? These issues are hardly new - they have arisen with textbooks and curriculum reforms of various sorts.55 But the (seeming) comprehensiveness that hypertexts make possible, and the largely invisible mechanisms of selection and organization that define their version of the truth, make them less susceptible to critical readings.56
Finally, the expense and sophistication of hypertext systems, and of the technology needed to run them, raises other sorts of questions for a democracy as well. Will citizens be able to access these information sources through publicly subsidized media, or will they be available only with a subscription price or other payment? Our society has had a remarkable commitment to the book, and to accessibility to books: even the smallest towns in the United States have some sort of public library. Will there be a similar public commitment to providing access to hypertexts? More to the point, for our purposes here, will public schools in different parts of the country have equal access to such resources, and the vastly enlarged learning opportunities they represent? Which students will also have access to such materials from their homes, and which ones not?
Are there ways to minimize these dangers?57 One might dream of making hypertext systems available to everyone; but their cost and complexity suggest that this will be difficult - and any subsidies, private or public, to help make them more affordable will often come with tacit conditions. One might dream of all hypertexts being completely interactive and dialogical, but this runs up against the realistic limits of knowledge and ability in most prospective readers, and introduces the risk of falling into the epistemological abyss of "everything being related to everything else." If all things are related to all things, and all relations are of equal value and importance, then even the knowledgeable user (let alone a novice) can become lost in an undifferentiated morass of information.
The dilemma here seems to be that as an organizing principle and as a potential educational, informational resource, hypertext either can provide too much information, and too loose a structure; or provide too selective a body of information, and too rigid a structure, rife with implicit judgments. Learners need heuristics, and in order to make some choices they need to have restricted options about others. But any set of heuristics will privilege particular structures of knowledge, and this exacerbates the stakes involved in deciding who will be selecting and organizing the information, who will be defining the criteria for relevance and relative importance of information, and who will be hard-wiring the most significant associative links into the system. It is essential that educators with an understanding of how different students learn, and with a sensitivity to questions of access and equity, play a role in the design of such materials. - for they will be developed (and marketed) with or without such input, to be sure. Yet it will take a significant change in self-perception for teachers to come to see themselves as designers of information systems - even though, in a sense, this is what they always were.
The poststructural, decentered state is one in which everything, prima facie, can have equal significance. There is something liberating about this, and something dangerous. Do we truly want a knowledge environment in which individuals can construct entirely personal and idiosyncratic ways of organizing information, with no eye to the fact that communities of culture and tradition have tended to prioritize things in certain ways rather than other ways? Does the leveling of all information nodes and the decentering of all organizing principles lead to more freedom or less?
On the other hand, as we have argued, creating explicit hierarchies and organizations of knowledge within a hypertext creates the potential for abuse. Hypertexts that are "hard wired," in which certain organizing structures cannot be overridden, or in which there are limits placed on the number and type of associated links the user can construct, or in which some information sources are restricted, instantiate a kind of hegemony: You can't ask that question. That information is not available to you. You have to define things this way to make any progress. You can't get there from here, and so on. Such statements would rarely be explicit in hypertexts; but they would inevitably exist, implicitly, in the choices underlying any particular hypertext.
To take full advantage of their potential, then, hypertexts ought to be rich, complex, open, and flexible; yet this may have the effect of limiting their usefulness for all but the most skillful and knowledgeable users. Providing access to the greatest number of readers and involving them in this technology will require making access to hypertexts simple, intuitive, and affordable, requiring a reduction in its scope and number of links, and limiting the preconditions of knowledge required to navigate it. Yet this ease of use comes at the cost of comprehensiveness, and entails that a good deal of implicit structure and selectiveness be built in.
Educationally, then, any hypertext system should meet some minimum level of interactiveness - that users be allowed to create some links between informational nodes based on their own associative frameworks, and not be limited only to the choices of the designers of the document. If hypertexts are indeed an environment conducive to intellectual exploration, then making them more dialogical is important for education in at least two respects: First, hypertext that limits the number or kinds of associative links that can be made by the user conceals and institutionalizes the authorial decisions about content and form that have been made by the producer of the hypertext. This raises the potential for serious abuse, as we have discussed. Second, by restricting where the user may explore within a hypertext, without allowing the possibility of supplementing these associations with new discoveries, hypertexts restrict not only what to think, but, potentially, how to think. This squanders one of the major potential benefits of hypertext, which is to tolerate and support multiple styles of learning and meaning-making.
One can see already a split between the basic kinds of hypertexts being produced. In some, the information is bounded and fixed, and the associative pathways, though they can be explored to some extent in idiosyncratic ways, are established by expert views of the subject matter and the associations learners should make. Other hypertexts contain some structure, but also permit more open-ended exploration and flexible responses to the textual material presented, and some allow the modification and customization of the system as readers begin to develop and add their own associative links.
In this regard, designers of hypertext for educational purposes can learn a great deal from the designers of certain computer games, especially those that involve mazes, puzzles, and problem solving in contexts of limited information (for example, the game Myst). Anyone who has become engrossed in such games (or to pick a less whimsical example, exploring the World Wide Web) knows the challenges presented in finding one's way around an information environment in which some pathways are laid out and others need to be searched; in which relationships of significance and purpose need to be explored through trial and error; in which there might not be one correct solution or goal but several. These sorts of hypertextual games or game-like environments, we are suggesting, demonstrate the tremendous appeal that the processes of exploration, discovery, and connection-making have for learners - if we can be creative and clever enough to use it for educationally substantial purposes.
However, developing potentials within the technology is fruitless if learners do not have the capacities and opportunities to exploit them fruitfully. Learning to hyperread is as complex and challenging a task as learning to read in the first place (or, in different words, it is a very sophisticated kind of reading). Part of this process may involve unlearning certain habits associated with reading linear texts, or regarding them as useful only given certain circumstances and purposes. The most important of these changes will be a shift away from a consumer approach to reading to a co-producer approach to reading, and a shift in one's view of gaining knowledge, away from the passive reception of facts, and toward the active construction of understanding.
Moreover, it will be crucial to consider in this educational process how different readers - representing different gender, racial, or cultural groups - encounter and respond to different kinds of structured information environments, what kinds of distinct barriers they experience, and what kinds of interactions might benefit their learning. There is no reason to expect that any particular form of hypertext will be suitable for all; or that the technological learning environment generally will be equally familiar and approachable to everyone. There is evidence already that the use of technology for learning merely privileges further the groups who can exploit it fully, leaving others who for whatever reason do not or cannot feel as comfortable with the technology even further behind.
One might wish that we could make everyone knowledgeable and skillful enough with hypertext to be "co-authors" in the fullest sense, but this is no more practical than expecting all hypertexts to be completely open and dialogical. Most readers of hypertext will be "browsers," or at most "users." But perhaps we can seek to educate critical users, readers who will know enough to use the system to find what they are looking for, but will also know enough to realize that what they have found might not be all that there is to know. They can approach the hypertext as an important resource, but maintain some skepticism about its reliability. Critical users will need to understand that what someone else has selected, interpreted, and organized for them may offer a partial and distorted picture of things. Users at this level might not have the skills to diagnose and find what is missing, but at least they will have a skeptical eye toward what they have found, and remain open to the possibility that there may be more. Eventually a few of these critical users may develop the knowledge and initiative to become co-authors - people who can actually move around within the hypertext and set up shop for themselves, creating new knowledge, constructing personalized systems, etc. We might wish that all readers will learn to do this, but this is not likely (in a society where so many VCR's still register a flashing "12:00" as the current time). The level of "critical user" may define the greatest level of sophistication that we can expect most readers to achieve in their interactions with hypertext; yet this is itself a significant educational objective, requiring teachers to develop new skills and understandings themselves, and to be willing to open up certain "authoritative" sources, such as encyclopedias, for critical scrutiny.
At the opposite pole from the critical user is the "surfer," the browser who jumps from information node to node, or from experience to experience, without regard to creating meaningful connections between them. There is some reason to worry that this phenomenological orientation is becoming more widespread in society - not only in computer contexts, but in switching relentlessly between cable television channels, between samples of music, through pages of magazines, across snatches of conversation. The sheer volume and variety of information, and of choices among pathways for accessing it, begin by forcing us to skim alternatives quickly to select what is worth closer attention - which in itself can be a valuable skill - but can end up limiting attention span and fostering a lack of reflectiveness about the choices actually made.
One might envision situations in which the technology itself can help students to become more critical users. Hypertext can be used as a tool to teach students multiple strategies for problem solving and information acquisition. In this process there will be an important role for teacher guidance and modeling: students, for example, might follow a teacher through the hypertext, observing and learning how someone with experience searches, collects, and links information.58 In progressive classrooms, hypertext will allow teachers and students to focus more on the important learning processes of interpreting and organizing information, and less on the trivial acquisition of facts. Teachers could be involved more with "scaffolding" - engaging learners at early stages with explicit explanation and guidance, leading them through hypertexts, and then gradually removing these supports as the learners become more independent and comfortable with exploring on their own. While we do not favor visions of the future without teachers and classrooms, there is no doubt that some of this instruction (introductory guided tours through a hypertext, for example) can and should become integral parts of the design of hypertexts themselves. There is also little doubt that some students are already able to access as much or even more hypertextual information through computer links in their own homes as can be found in their schools., while others have no such access. As noted previously, the dangers this raises for a significant new form of educational inequality, one as serious and limiting for learners as illiteracy of other types, should be a pressing concern for any educator interested in opportunity and fairness. The skills and attitudes of being an effective co-author or critical user should not become the special domain of certain groups and not others; indeed, part of the very capacity of reading hypertexts critically, diagnosing their distortions, biases, and gaps, will require that they be studied by readers who see the world differently from those who designed the hypertexts.
We stand at a crossroads where the very technology that offers the means to broaden access to liberating knowledge is just as likely to promote a hegemonic concentration, in the hands of the few, of the "means of production" for shaping and organizing information. Hypertext makes possible some radically new educational possibilities. If we want to play a role in shaping these possibilities, and to influence them along progressive lines, educators will need to develop new skills in information design and interpretation. We will need to take the lead in helping others to develop these skills. And we will need to initiate serious reflections upon the social, moral, and epistemological consequences of technology's influences on teaching and learning. We hope that this essay can help to spark such reflections.*
1 The sequential ordering of authors is itself in part an artifact of a particular kind of traditional text. The authors of this manuscript have shared the work and responsibility for writing this manuscript equally; in a hypertextual environment, it would be much easier to reflect this in a nonlinear, nonhierarchical fashion. (See also Cliff McKnight, Andrew Dillon, and John Richardson, Hypertext in Context (Cambridge: Cambridge University Press, 1991), p. ix.; and Paul Delaney and George P. Landow, "Hypertext, hypermedia, and literary studies," in Paul Delaney and George P. Landow, eds, Hypermedia and Literary Studies. (Cambridge, MA: MIT Press, 1991), pp. 14-17).
2 Like many authors, we will use the term "hypertext" inclusively, referring as well to so-called "hypermedia" environments. Hypermedia refers to a hypertext system that links various media (pictures, sound, etc.) as well as written text per se. While the issue of multiple media as sources of information introduces a variety of important issues, especially for questions of learning and alternative learning styles, all of these media can be considered "texts" in a broad sense of the term, and when organized as hypertexts, they encounter many of the same basic issues as purely written hypertexts.
3 The plain text and italicized text in the manuscript do not correspond directly to the two authors' contributions.
4 Several good introductory books on this subject also adopt elements of a hypertext model in the way they present information. For example, see Jay David Bolter, Writing Space: The Computer, Hypertext, and the History of Writing (Hillsdale, NJ: Erlbaum, 1991), which is available in both printed and computer disk versions; Robert E. Horn, Mapping Hypertext (Lexington, MA: Lexington Institute, 1989); and David H. Jonassen, Hypertext/Hypermedia (Englewood Cliffs, NJ: Educational Technology, 1989). Other helpful introductory material is available from George P. Landow, Hypertext: The Convergence of Contemporary Critical Theory and Technology (Baltimore: Johns Hopkins University, 1992); a special issue of Educational Technology, Vol. 28 No. 11 (1988), edited by Gary Marchionini; Jakob Nielsen, Hypertext and Hypermedia (New York: Academic Press, 1990); Ben Shneiderman and Greg Kearsley, Hypertext Hands-on! (New York: Addison-Wesley, 1989); and McKnight, Dillon, and Richardson, Hypertext in Context.
5 Peter Holland, "Authorship and collaboration: The problem of editing Shakespeare," in Warren Cherniak, Caroline Davis, and Marilyn Deegan eds., The Politics of Electronic Text (Oxford: Office of Humanities Communication, 1993), p. 20.
6 Vannevar Bush, "As we may think," Atlantic Monthly, Vol. 176 No. 1 (1945), pp. 101-108.
7 Quoted in J. Conklin, "Hypertext: An introduction and survey," IEEE Computer, Vol. 20. No. 9 (1987), pp. 17-41.
8 Although Nelson first used the term much earlier, this definition comes from his Literary Machines (Swarthmore, PA: Self-published, 1981), p. 2.
9 Julio Cortazar, Hopscotch (NY: Avon Books, 1966). Another, more recent book, with a strongly "hypertextual" look and feel is Avital Ronell, The Telephone Book (Lincoln, NE: University of Nebraska Press, 1989).
10 In hypermedia, we may not even be talking about words or pages anymore, but pictures, music, images, etc.
11 See Cherniak, Davis, and Deegan, The Politics of Electronic Text.
12 For a discussion of several of these analogies, see J. Conklin, "Hypertext: An introduction and survey," IEEE Computer, Vol. 20. No. 9 (1987), pp. 17-41, Robert E. Horn, Mapping Hypertext (Lexington, MA: Lexington Institute, 1989), and Cliff McKnight, Andrew Dillon, and John Richardson, Hypertext in Context (Cambridge: Cambridge University Press, 1991).
13 John Slatin, "Reading hypertext: Order and coherence in a new medium." College English, Vol. 52 No. 8 (1990), 870 - 883.
14 For the impact on these new technologies on changing conceptions of the library, see Gregory T. Anderson, "Dimensions, context, and freedom: The library in the social creation of knowledge," and Kathleen Burnett, "Multimedia and the library and information studies curriculum," in Edward Barrett, ed., Sociomedia: Multimedia, Hypermedia, and the Social Construction of Knowledge (Cambridge, MA: MIT Press, 1992), pp. 107-124 and 125-139.
15 Gilles Deleuze and Felix Guattari, "Rhizome," in On the Line (New York: Semiotext(e): 1983), pp. 1- 65. Thanks to Zelia Gregoriou for this reference.
16 Deleuze and Guattari, "Rhizome," pp. 11, 47-49.
17 And only with a computer can these linkages be made conveniently with multimedia sources.
18 C. Carr, "Hypertext: A new training tool?" Educational Technology, Vol. 28 No. 8 (1988), 7-11. Delaney and Landow, Hypermedia and literary studies.
19 Michel Foucault, "What is an author?" Textual Strategies: Perspectives in Post-structuralist Criticism, ed. Josue Harari (Ithaca, NY: Cornell University Press, 1979), pp. 141-160. This shift in perspective may also have significant implications for changing views of copyright and intellectual ownership. See Nicholas C. Burbules and Bertram C. Bruce, "This is not a paper," in review.
20 See David Jonassen's helpful discussion of hypertext and schema theory, in Hypertext/Hypermedia, p. 23. See also: David Chen, "An epistemic analysis of the interaction between knowledge, education, and technology," in Barrett, Sociomedia: Multimedia, Hypermedia, and the Social Construction of Knowledge, pp. 161-173; M.C. Linn, "Hypermedia as a personalized tool for knowledge organization," presented at the American Educational Research Association meetings (April 1991), in Chicago, Illinois; and Rand J. Spiro, Richard L. Coulson, Paul Feltovich, and Daniel K. Anderson, "Cognitive flexibility theory: Advanced knowledge acquisition in ill-structured domains," In Tenth Annual Conference of the Cognitive Science Society (Hillsdale, NJ: Erlbaum, 1988), pp. 375-384; Rand J. Spiro and J. Jehng, "Cognitive flexibility and hypertext: Theory and technology for the linear and multidimensional traversal of complex subject matter," in Spiro, R. J. & Jehng, J., eds., Cognition, Education, and Multimedia: Exploring Ideas in High Technology. Hillsdale, NJ: Erlbaum. (1990), pp. 164-205; and Rand J. Spiro, Paul Feltovich, Michael J. Jacobson, and Richard L. Coulson, "Cognitive flexibility, constructivism, and hypertext," Educational Technology (May, 1991), pp. 24-33.
21 Slatin, "Reading hypertext," p. 873.
22 Delaney and Landow, "Hypertext, hypermedia, and literary studies," p. 10.
23 On "bricolage," see Claude Levi-Strauss, The Savage Mind (Chicago: University of Chicago Press, 1966), pp. 16-37.
24 For a discussion of this problem and its consequences, see Bernard Williams, "The Riddle of Umberto Eco," New York Review of Books, February 2, 1995, pp. 33-35.
25 The best single source on this relationship is Landow, Hypertext; see also Delaney and Landow, "Hypertext, hypermedia, and literary studies." Additional discussions of the affiliation between hypertext and postmodernism are Norman N. Holland, "Eliza meets the postmodern," EJournal, Vol. 4 No. 1 (1994) and Bolter, Writing Space; see also a discussion of Bolter's book in EJournal by Joe Amato and Doug Brent: EJournal, Vol. 1 No. 2 (1991) and Vol. 1 No. 2-1 (1991).
26 Delaney and Landow, "Hypertext, hypermedia, and literary studies," p. 18.
27 Slatin, "Reading hypertext."
28 Umberto Eco, A Theory of Semiotics (Bloomington, IN: Indiana University Press, 1976), pp. 69 ff.
29 Jay David Bolter, "Topographic writing: Hypertext and electronic writing," in Delaney and Landow, eds, Hypermedia and Literary Studies. (Cambridge, MA: MIT Press, 1991), pp. 105-118; see also Bolter, Writing Space; Cliff McKnight, John Richardson, and Andrew Dillon, "The authoring of hypertext documents," in Ray McAleese, ed., Hypertext: Theory into Practice (Norwood, NJ: Ablex, 1989), pp. 138-147.
30 Stanley Fish, Is There a Text in this Class? The Authority of Interpretive Communities," (Cambridge, MA: Harvard University Press, 1980, pp. 322-327.
31 Roland Barthes, S/Z: An Essay (New York: Hill and Wang, 1974). For another discussion of this example, see George P. Landow, Hypertext: The Convergence of Contemporary Critical Theory and Technology (Baltimore: Johns Hopkins University, 1992).
32 David Lodge, Small World, NY: Warner Books, 1984.
33 Jorge Luis Borges, "Kafka and His Precursors," in Labyrinths (NY: New Direction, 1964), p. 201. Thanks to Punya Mishra for suggesting this reference.
34 See Jacques Derrida, Glas (Lincoln: University of Nebraska Press, 1986), in which Derrida publishes two essays, on apparently unrelated topics, side by side.
35 For a discussion of issues in electronic publishing, see Nicholas C. Burbules and Bertram C. Bruce, "This is not a paper," in review.
36 Delaney and Landow, Hypermedia and Literary Studies; Slatin, "Reading hypertext"; T. Byles, "A context for hypertext: Some suggested elements of style." Wilson Library Journal, Vol 63 No. 3 (1988), 60 - 62.
37 Delaney and Landow, Hypermedia and Literary Studies, p. 19.
38 David Jonassen, "Designing structured hypertext and structuring access to hypertext." Educational Technology, Vol. 28 No. 11 (1988), pp. 13-16.
39 See also Slatin, " Reading hypertext," p. 875, and Ray McAleese, "Navigation and browsing in hypertext," in Ray McAleese, ed., Hypertext: Theory into Practice (Norwood, NJ: Ablex, 1989), pp. 6-44.
40 Slatin, "Reading hypertext," p. 875.
41 See also Horn, Mapping Hypertext, pp. 11, 26-27.
42 Shneiderman and Kearsley, Hypertext Hands-on!, quoted in Henrietta Shirk, "Cognitive architecture in hypermedia instruction," in Barrett, ed., Sociomedia: Multimedia, Hypermedia, and the Social Construction of Knowledge, p. 81.
43 Jonassen, Hypertext/Hypermedia, quoted in Shirk, "Cognitive architecture in hypermedia instruction," pp. 82, 85-86.
44 Shirk, "Cognitive architecture in hypermedia instruction," p. 86.
45 With currently emergent technology, it may be able to create "intelligent agents," tailored search procedures that can find, select, and organize information from multiple sources, according to specifications identified in advance.
46 Ted Nelson was probably the first to advocate this; see Jonassen, Hypertext/Hypermedia, p. 22. See also Nielsen, Hypertext and Hypermedia, p 13.
47 See Deborah Edwards and Lynda Hardman, "'Lost in hyperspace': Cognitive mapping and navigation in a hypertext environment," in McAleese, ed., Hypertext: Theory into Practice, pp. 105-125; N. Hammond and L. Allinson, "Extending hypertext for learning: An investigation of access and guidance tools," in A. Sutcliffe and L. Macaulay, eds., People and Computers V. (Cambridge: Cambridge University Press), 1989); and P. Brown, "Do we need maps to navigate around hypertext?" Electronic Publishing, Vol. 2 No. 2 (1989), pp. 91-100. See also Horn, Mapping Hypertext, pp. 50-59, 150-159; Jonassen, Hypertext/Hypermedia, pp. 41-45; and Nielsen, Hypertext and Hypermedia, pp. 127, 143 - 162.
48 David Gelernter, "Unplugged," The New Republic (September 19 & 26, 1994), pp. 14-15.
49 One of many possible examples of this tone can be found in Holland, "Authorship and collaboration," p. 22.
50 See several essays in Cherniak, Davis, and Deegan eds., The Politics of Electronic Text, including Ian Small, "Text-editing and the computer: Facts and values," pp. 25-30; Marcus Walsh, "The fluid text and the orientations of editing," pp. 31-39, and Michael Leslie, "Electronic editions and the hierarchy of texts," pp. 41-51.
51 Dale Spender, "Feminism does not compute: The Computer Age - Implications for feminism," a talk given at the University of Illinois, Urbana/Champaign, April 17, 1991. For an example of an avowedly feminist approach to text design, see: Kathryn Sutherland, "Challenging assumptions: Women writers and new technology," in Cherniak, Davis, and Deegan eds., The Politics of Electronic Text, pp. 53-67. Also see, "Who controls the text?" chapter 6 of Landow, Hypertext.
52 M.E.D. Koenig, "Linking library users: A culture change in librarianship," American Libraries, Vol. 21 No. 9 (1990) 844-845, 847, 849.
53 Michael W. Apple, Teachers and Texts (New York: Routledge, 1986).
54 But not impossible; for a wonderful example of using critical readings of hypertexts to decode their hidden biases, see Bob Peterson, "Bias and CD-ROM encyclopedias," Rethinking Schools, Vol. 9 No. 1 (1994), pp. 6-7.
55 A more general discussion of educational implications of hypertext can be found in Landow, Hypertext, chs. 5 and 6.
56 The authors wish to credit the excellent suggestions and criticisms of several colleagues: Chip Bruce, Zelia Gregoriou, Craig Howley, Marcia Linn, Punya Mishra, Eugenie Potter, Pamela Salela, Leslie Sassone, Evelyn Shapiro, and Rand Spiro. We especially want to thank Ralph Page, Walter Feinberg, and the reviewers of Educational Theory for their comments and suggestions during the editorial process.