This paper was refereed by the Journal of Electronic Publishing's peer reviewers.

This article was originally published as hypertext, and is best viewed in its original format, which is available on our website. Because today's technology does not allow us to guarantee its long-term availability in the original format, we display here the preservation-quality copy, which is displayed in linear form.

 
Narrative languishes as the unsolved puzzle of hypertext.

Linked presentation as we know it today — a navigation menu, a table of contents, a list of frequently asked questions (FAQ) — suits informational material such as technical manuals, government documents, and most scientific research papers. These presentation formats do little to enhance narrative forms, however.

Most discussion of online narrative — and most experimentation — has centered on fiction (Coover, 1993, 2000; Minganti, 1996) and literary studies (Landow, 1992; Lavagnino, 1997). Journalism narratives, especially long-form journalism, are overdue for attention.

Steven Johnson, writing in 1997, noted that "The great preponderance of Web-based writing is unapologetically linear. Almost all journalistic stories are single, one-dimensional pieces, articles that would be exactly the same were they built out of ink and paper instead of zeros and ones.... The individual articles themselves rarely offer any navigational options at all" (p. 128).

A new process for writing is needed if online narratives are ever to be compelling. We believe it is not only desirable but also necessary to move journalistic, nonfiction, and even scholarly writing in a direction made possible by hypertext. In this paper we argue that hypertext forms improve the reader's experience of reading, and therefore, hypertext is better for telling particular nonfiction stories.

Rather than deriving the pieces of a hypertext by disassembly, the writer must defer the practice of assembling the text until after all the components are complete — this practice creates texts that can be adapted as easily to online as to print (whereas texts created for print resist adaptation to hypertext).

The process of structuring these components into a multilinear narrative requires the writer to impose order and make decisions about the threads that will be available to readers. Decisions regarding linking and omission give rise to questions about the author's responsibilities to the story and to objectivity.

If the reader's experience of agency is heightened in hypertext narratives, does it follow that the writer relinquishes authority over the text? That's the control paradox: We contend that the author gives up very little — perhaps nothing at all.

When hypertext is better

As more readers become accustomed to hypertext, they may feel unsatisfied when scrolling through linear stories formatted as if for print media.

Browsing through nonlinear hypertexts is not without drawbacks, however. Users often describe a disorientation: They repeatedly lose their place or forget to follow up on their intentions; they wonder if they are missing anything (Foss, 1989).

Readers who have spent their lives "in love with books" may always feel ambivalent toward hypertext (Murray, 1997). But their mixed feelings do not license writers to ignore the ways hypertext improves upon current forms for delivering information.

Hypertext enables writers to skirt issues raised by the use of predefined writing formulas such as the inverted pyramid of traditional news stories which present facts in decreasing order of importance, the narrative path fixed and loaded with value judgments. Some newspaper readers stop after a few paragraphs, satisfied that they've read the "most important" information. But what they got is the part of the story the writer believes is most important. Other readers skip through an article, trying to find a perspective relevant to them.

In contrast to linear stories with distinct beginnings, middles, and endings, hypertexts generally have multiple possible entry points, many internal threads, and no clear ending (Murray, 1997). This makes hypertext especially suitable for stories with numerous components. Unlimited storytelling space makes it so no component need be omitted in the process, which can lead to more objective journalism.

With a good hypertext, readers can foreground their own perspectives by first clicking links to the parts of the story they see as more important, then reading the remaining parts in any order (Fredin, 1997).

In a unilinear story (i.e. print tradition), readers, reading conventionally, see first what the writer thinks matters most. In a well-crafted hypertext, readers immediately select what matters most to them. They spend less time sifting through facts as filtered through the writer's mind. Readers' choices allow them to avoid a form of information overload brought on by the choices of the writer.

Components defined

When an article is broken apart, the reader can move around at will within the text, choosing an order for reading that suits her or him.

Two key aspects of these separate components:

1. Each component is tightly focused on a single idea, event, description, or problem.
2. No component of an article substantially repeats anything stated in another component in the same article.

Each piece of an article will average about 250 words. Some pieces, or components, will be longer (up to about 350 words at most) and some will be shorter (usually no less than 150 words). These lengths are not arbitrary; an examination of most newspaper writing, and many other nonfiction texts, reveals that journalists (and others) already write naturally in blocks of similar size.

(In seminars and courses on writing for the Web, we have given participants a printed copy of a traditional long-form news story or essay and asked them to mark the "break points" where the time, place, or subject changes significantly. Participants largely agreed on all the break points, most of which fall at the intervals noted above.)

Creating an online text made up of separated components has been called "chunking" or "nonlinear narrative" (Nelson, 1990; Murray, 1997; Nielsen, 2000, esp. pp. 112-15).

"Broken apart" is perhaps a poor metaphor; it implies a whole that has been damaged. We assert that components ideally are written discretely as components from the beginning of the process. Extracting components from a pre-existing unilinear text not only proves to be equally (or more) difficult but also appears to produce inferior components.

Building hypertexts: Process

The writer begins by making a list of all the ideas he or she wants to include and writes each of these separately as a component.

Then the writer creates structure. Sometimes the writer creates structure before he or she has finished writing — that is okay. Then the writer writes some more. Revisions in the structure will be necessary, and afterward, some or all components will also need rewriting. Finally, the writer writes the links that instantiate the structure. These tasks typically overlap.

Murray (1997) describes the writer's work as procedural authorship, a process that involves building structure, writing texts, and establishing relationships among the texts. Procedural authorship includes:

1. Writing components. Once written, components usually require revision as structure develops.
2. Determining sequences (order) for the article components. The writer decides on possible reading orders for all components.
3. Building the possible orders into the hypertext with links.

Murray (1997) does not extract linking as a key and separate step, but we identify linking as a significant, distinct part of the process. Component-writing runs parallel to structure-building, and linking runs parallel to these. The writer builds structure with components that stand on their own; components are never merged (as in print) but only linked.

Linking marks a critical difference between authoring a unilinear text and building a hypertext with multiple reading orders. Linking cannot be ignored. Links give hypertexts their flex, enabling readers to follow their preferred paths through articles.

The authors of this article worked according to these steps, and we will be the first to say it: This process is labor intensive. And yet, additive construction over time may give depth and breadth to hypertext structures. There is no reason to consider the "finished" hypertext fixed and unalterable.

The control paradox

Writers (and some editors) fear hypertext because they hope to control the reader's experience. Control seems important because without it, the reader may not get the message, or may get the wrong message.

Readers, however, have never been constrained to follow the order imposed by a writer. On the printed page, the reader's eye has always been free to bounce and skip, to return or not. Readers bring with them all their past experience and knowledge, which the writer cannot control (or even make many assumptions about); these determine the readers' interests.

We defend two arguments here:

1. The writer does not give up control in hypertext.
2. The reader has always had a large degree of control.

Pang (1998a) describes how authors for the electronic Encyclopedia Britannica become more important as they take on responsibilities to integrate their articles with other media and with the work of other authors. Britannica editors face new concerns about the interrelationships of different articles and the reader's experience in moving from article to article.

Rather than giving up control of these interrelationships, the authors evaluate possible connections, making choices about the links available to readers. In the encyclopedia, the perception of the discrete, stand-alone article is becoming obsolete (Pang, 1998a).

Readers experience any article in a personal way that envelopes them and the text in a singular universe, a universe where many other texts interact with the one being read, but where the reader is the only actual person. The author does not exist in that universe, except as a shadow cast by the text itself. During the time when reading takes place, all control belongs to the reader (Barthes, 1968).

The writer maintains control over the act of writing and creating in hypertext, just as in the act of writing for printed media. The writer chooses the words, examples, anecdotes, facts, scenes, characters. The writer imposes order in hypertext just as in print. When the writing is finished, however, the writer's control ends.

Online as on paper, with pixels as with ink, the reader controls the experience of reading.

Journalism: "Objectivity"

Will hypertext compromise journalism? That question rests on the possibility that stories told in the manner described here will be less fair, less credible — even less factual — than traditional print journalism.

Journalists often use the term "objectivity" as a shorthand to represent all matters concerning fairness, credibility, and accuracy. Many argue that "objectivity" went out of fashion in the mid-20th century as intellectuals conceded that all people (including journalists) have biases, sometime after Merleau-Ponty rejected the Husserlian construct of bracketing one's lived experience. Journalistic objectivity, nevertheless, retains currency as an ideal and sometimes as a competitive weapon (Mindich, 1998).

To approach the question of objectivity in hypertext, it will be useful first to determine the journalistic practices in which objectivity might apply. Gauthier (1993) reasoned that:

• Objectivity can be applied only to "straight news reporting" (e.g., not to opinion pieces, advocacy journalism, analysis, or commentary);
• Objectivity does not apply to news gathering (a news story "cannot ... be an exact replica" of events); and
• Objectivity does not apply to the "formal and material constraints affecting the press" — while such constraints can and do hamper journalists, they do not make objectivity impossible.

Having eliminated those aspects of the practice of journalism, there is one more to be considered, which concerns the basic relationship between the journalist and the facts reported — "the way the journalist processes information" (Gauthier, 1993).

The practices of the journalist in selecting (and omitting) and presenting (or ordering) information — the practices of authoring — constitute the situation in which objectivity might be applied.

According to press critic Jay Rosen, "What is insidious and crippling about objectivity is when journalists say: 'We just present you with facts. We don't make judgments. We don't have any values ourselves.' That is dangerous and wrongheaded" (Glaberson, 1994).

The structure of a news story is largely determined by the values of the journalist and journalism institutions; these values affect the decisions made in the act of authoring. Because a hypertext structure is multilinear — a collection of linked components — the standards of fairness and balance, and implications of bias and influence, will be manifest in the links themselves (Fredin, 1997).

Authoring: Choices in omission

In print, the writer inevitably chooses what to cut from an article. Online, however, writers discover they have to add more — not by their editors’ standards, but from a duty to tell the full story.

The endless writing space online has created a profound new ethical obligation for writers of factual articles: more complete reporting. This obligation requires thoughtful consideration of what gets left out.

Traditional publishing routines require omission (Koch, 1991). Limited writing space often interferes with a writer's ability to include all information relevant to a given article.

Newsroom routines have hindered the production of online news, making online journalism an imperfect mirror image of its print counterpart; however, some reporters and editors are learning that articles "acceptable and accessible" in one medium are not necessarily so in others (Martin, 1998, p. 65).

Online, constraints on article-length do not exist: Endless space shocks writers with the possibility of adapting to new routines.

The writer knows cyberspace will accommodate a more complete version of the article — and that hypertext makes it possible to show all the parts of a story in relationship to one another. Before deciding what to scrap, the writer weighs how omitting any one part alters the whole.

For example: While reporting on a public advisory board's 5-4 decision to recommend funding for a new bike trail, a writer might interview the nine board members and several citizens living near proposed trail areas.

For print, the writer probably would present the arguments through quotes from two or three board members and one or two concerned citizens. Space permitting, he or she might write up each interview separately, but it is unlikely. Presenting each interview as a discrete unit would not make sense in a linear article, and there would not be enough space.

Online, each interview could be a component. Which perspective(s) could the writer fairly omit?

Authoring: Choices in order

Choices a writer makes in ordering a hypertext limit the choices of the reader. The writer decides what part of the story readers see first, as well as what links readers can follow from there. This principle applies to all pages in the hypertext; the writer chooses which parts of the article are accessible from other parts of the article and which are not.

Linking decisions create different reading orders for the same story. Different people would link the same texts differently, depending on their knowledge of the subject and their idea of good hyperlinking (Pang, 1998b).

The ordering process includes:

• Evaluating the major threads.
• Selecting the ideas that matter most.
• Considering what the majority of readers will be curious about or will need to know more about.

Once the writer selects the primary threads, all the others become secondary to those. Depending on the size of the hypertext, tertiary and even more "removed" threads may also be constructed. The reader will not discover these threads until they become relevant.

The writer's control of hypertext construction does not make the article the same, ultimately, as a linear text. A linear article offers one thread defined by the writer. The writer structures the article one way with a beginning, middle, and ending. This choice of story order gives the writer the official "final word."

In a nonlinear article, the writer establishes "a structure of possible structures" for the article (Bolter, 1991). Depending on how the writer links components, multiple possible reading orders emerge. Choosing from these, the reader defines a path through the article, deciding on his or her own final word.

But the writer, through linking, establishes the orders in which possible "final" words are available in the first place. Choices made in ordering a hypertext represent control reserved for the writer.

Links, used wisely

A signpost on a path offers the reader a choice: To go or not to go. The link represents not only a connector, but also a division, a point where the path forks (for an analysis, see Harpold, 1991b).

A word, a phrase, a sentence, an image — whatever its form, the link provides a portal to another place. The link may give the reader a clear idea of what that place will be like, or it may seem very obscure (depending on the link itself and on the individual reader). The reader asks, "Should I risk it?" Each time readers follow a link, they face disappointment, surprise, reward, disorientation.

The reader's expectations of any link are difficult to gauge, because most Web users have had prior experience with links — at many different sites, good and bad. Who can say with certainty what "next" or "back" might mean?

Inventing connections: The writer does this by determining what structure the hypertext will have, by building a hierarchy of threads, and finally, by creating the links.

The writer decides:

• Which links should appear on which components?
• What will the links look like? (Images or text? Underlined or not?)
• If the links incorporate text, what words will be used?
• Will the links be embedded within paragraphs or listed separately?
• If the components have distinct titles and headings, what will be the relationship of the link text to those?

When they have been deliberately located and carefully crafted, links create coherence (Johnson, 1997). If a reader finds that links confuse or disappoint, lead to irrelevant material, or fragment the text without reason — look to the decisions made by the writer.

Four bad links

See also Ten link rules.

Ten link rules

1. Give the user choices: Offer more than one link on each page.
2. Do not use similar phrases to link to different pages (i.e., "Who We Are" linked to a staff photo page and "About Us" linked to a company profile). Do not create False Twin Links (see Four bad links).
3. If you must link to the same thing more than once on a single page, use very similar text or the same graphic for each of the links. Do not create Non-Identical Twin Links (see Four bad links).
4. Unless you are creating a comprehensive directory, do not offer users a lot of similar options; be selective. Do an editor's job: Choose the best and eliminate the rest.
5. Do not give users too many options, even dissimilar ones. Again — do an editor's job by eliminating the less important links.
6. Do not hide or "bury" links to pages to which many users will want access (such as "How to Order" or "References"). Anticipate the users' goals and desires, and prioritize.
7. Avoid irrelevant, extraneous, or unnecessary links. They burden users and make your site seem less useful.
8. Do not send users away from your site without a good reason. The page you send them to should be relevant, excellent, and not like anything you could (or do) offer on your own site.
9. Write the link text to give a reasonable expectation of what the link will deliver. Do not create Mystery Links or Trick Links (see Four bad links).
10. Never use the phrase "click here." It does not tell users anything.

Threads or pathways

A "thread" is a path that the reader follows through an article. A typical written-for-print article has a single thread, from beginning to end, which the writer expects the reader to follow. The construction of this thread lies at the core of the craft of writing.

The absence of a unilinear thread marks the most obvious difference between hypertext and the typical printed text. The responsibility of constructing threads remains with the writer, but the writer acknowledges that hypertext form requires multiple threads — not just one.

If the writer provides access to all the components from the very start of the hypertext (by linking them there), and maintains that same complete access in each component, then the reader can construct a thread almost at will — whatever path the reader takes will become the thread for that reader.

Making all the components equally available is practical in smaller hypertexts. With a small number of components (six or fewer), the reader can evaluate the links to all of them and choose one, without too much difficulty. In such a hypertext, all the threads are reader-built threads.

With a larger number of components, the writer must construct a hierarchy so that the reader will not be burdened with confusing decisions about where to go next. This hierarchy opens particular pathways (or threads) and closes off others. The possible pathways are author-built threads.

Reader-built threads

A brief hypertext might be considered "thread-independent": If no components are ever hidden, then readers are free to choose any order for reading. This is the case when all the links in the entire hypertext are available in all components.

The only order that the writer might impose in a brief hypertext would affect links in a list. A reader may be more likely to select the link at the top of a list and to follow the list order from top to bottom.

If the order of links in the list rotates or cycles so that a different link is always at the top when the reader opens a new component, the reader may assume that the link at the top of the list leads to the "next" page. However, this cycling may disorient the reader by making the link list appear to be unstable.

A list of links does not constitute an author-built thread; the reader still can choose any link on the list. The writer has placed no restrictions on the reader's movement within the hypertext if all links are always available.

Author-built threads

In a larger hypertext (with seven or more components), the writer confronts the need to construct threads. Such hypertexts can be considered "thread-dependent" or "deliberately threaded."

Readers experience some strain when they face a list of more than six links to different components — having more choices makes it harder to choose. (Remember the last time you were in a ice cream shop with dozens of flavors?) The chance of error (or disappointment) increases because the differences between items on the list are harder to infer.

If the writer imposes a hierarchy, the material becomes more manageable for the reader. The hierarchy hides or buries certain components that the writer deems less vital. The writer ensures that the reader will choose one of the more important components to begin with if the writer has limited access to a select subset of the total number of components in the hypertext.

(The hypertext you are reading now include[d] [in its original format] a "top page" where only four components [were] available.)

Reader-built threads also occur in such larger hypertexts, because readers still get to choose from several options for where to go next, even if some options are hidden from them in some places.

The reader's experience

Readers moving from one component of a hypertext to another will experience agency — if the writer has linked well.

Murray (1997) defines agency as the power of readers to take meaningful action and see the results of their decisions and choices. She notes that computer users "expect" to feel agency when clicking on a file. When the file opens, that power satisfies the user.

Readers do not usually expect to find agency in a narrative environment (Murray, 1997). But with hypertext, the writer can allow readers to experience the agency they may "expect" when interacting with a computer screen. When the writer allows reader agency by writing good links, the hypertext enables readers to come to their own conclusions about a story.

Links create associations; the connections are deliberate, not random. This marks a crucial difference between clicking through Web pages and surfing through TV channels (Johnson, 1997). As always, the writer cannot guarantee which associations the reader will make.

The writer also cannot predict which links the reader will follow. This means every link counts. Some readers feel satisfied to click one link and read one component. Other readers feel some compulsion to make certain they have "read everything" and otherwise feel dissatisfied, perhaps cheated; they may resent the resulting increase in their information overload.

Our theory is that the reader should, at some point, feel satisfied — whether or not she has read everything. Feeling satisfied, a reader won't care if there are components she did not read. She got enough, whatever "enough" means for her.

The reader's experience always will be formed in relation to the writer's link decisions. Out of the resulting sense of satisfaction or frustration, trust in (or validation of) the hypertext rises or fails. Did the writer's link seem to make sense? Did the reader say "Huh? What?" Or did the reader say "Oh, nice!" or "What a crock!" (See descriptions of Four bad links.)

Readers have the final word. Their agency cannot be denied.

Mindy McAdams (author bio) [email protected]

Stephanie Berger (author bio) [email protected]

References

Barthes, R. 1968. The death of the author. In Image - Music - Text, 142-48. New York: Hill and Wang.

Bolter, J. D. 1991. Writing space: The computer, hypertext, and the history of writing. Hilldale, N.J.: Lawrence Erlbaum Associates.

Conklin, E. J. 1987. Hypertext: An introduction and survey. IEEE Computer, v.20, n.9: 17-41. [Online]. Formerly: http://www.ai.univie.ac.at/~paolo/lva/vu-htmm1998/html/conklin87/Conklin87.html

Coover, R. 1993. Hyperfiction: And hypertext is only the beginning. Watch out! [Online: Hyperfiction: Novels for the computer. The New York Times, 29 August, sec. 7 (Book Review), 8ff.] Available: http://www.nytimes.com/books/98/09/27/specials/coover-hyperfiction.html

Coover, R. 2000. Literary hypertext: The passing of the Golden Age. In FEED, February. [Online]. Formerly: http://www.feedmag.com/document/do291.shtml

Engelbart, D. C. 1962. Augmenting human intellect: A conceptual framework. AFOSR-3223 Summary Report. Menlo Park, Calif.: Stanford Research Institute, October. [Online]. Available: http://www.bootstrap.org/augdocs/friedewald030402/augmentinghumanintellect/ahi62index.html

Fredin, E. 1997. Rethinking the news story for the internet: Hyperstory prototypes and a model of the user. Journalism and Mass Communication Monographs, 163, 1-47.

Foss, C. 1989. Tools for reading and browsing hypertext. Information Processing and Management, 25: 407-418. [doi: 10.1016/0306-4573(89)90068-X]

Gauthier, G. 1993. In defence of a supposedly outdated notion: The range of application of journalistic objectivity. Canadian Journal of Communication, v.18, n.4: (unpaginated). [Online]. Available: http://www.cjc-online.ca/viewarticle.php?id=201&layout=html

Glaberson, W. 1994. Fairness, bias and judgment: Grappling with the knotty issue of objectivity in journalism. The New York Times, 12 December, sec. D: 7.

Harpold, T. 1991a. The contingencies of the hypertext link. Writing on the Edge, v.2, n2: 126-37. [Online]. Available: http://www.newmediareader.com/cd_samples/WOE/Harpold.html

Harpold, T. 1991b. Threnody: Psychoanalytic digressions on the subject of hypertexts. In Hypermedia and Literary Studies, edited by Paul Delany and George P. Landow, 171-84. Cambridge, Mass.: MIT Press.

Johnson, S. 1997. Interface culture: How new technology transforms the way we create and communicate. New York: Basic Books.

Koch, T. 1991. Journalism for the 21st century: Online information, electronic databases, and the news. Westport, Conn.: Praeger.

Landow, G. P. 1992. Hypertext: The convergence of contemporary critical theory and technology. Baltimore, Md.: Johns Hopkins University Press.

Lanson, J. 2000. August 24. It's time to make the Web an equal partner. Online Journalism Review, 24 August (unpaginated). [Online]. Available: http://www.ojr.org/ojr/business/1017963559.php

Lavagnino, J. 1997. Reading, scholarship, and hypertext editions. Journal of Electronic Publishing, v.3 n.1 (September). [doi: 10.3998/3336451.0003.112]

Martin, S. E. 1998. How news gets from paper to its online counterpart. Newspaper Research Journal, v.19, n.2: 64-73.

Mindich, D.T.Z. 1998. Just the facts: How "objectivity" came to define American journalism. New York: New York University Press.

Minganti, F. 1996. Updating (electronic) storytelling. Technology and American Culture, a special issue of Amerikastudien (American Studies), v.41, n.3: 417-30.

Murray, J. H. 1997. Hamlet on the holodeck: The future of narrative in cyberspace. Cambridge, Mass.: MIT Press.

Nelson, T. H. 1990. Literary machines 90.1. Sausalito, Calif.: Mindful Press (originally published in 1982).

Nielsen, J. 2000. Designing web usability: The practice of simplicity. Indianapolis: New Riders.

Nielsen, J. 1995. Multimedia and hypertext: The internet and beyond. Boston: AP Professional.

Pang, A. S-K. 1998a. The work of the encyclopedia in the age of electronic reproduction. First Monday, v.3, n.9. [Online]. Available: http://www.firstmonday.org/issues/issue3_9/pang/

Pang, A. S-K. 1998b. Hypertext, the next generation: A review and research agenda. First Monday, v.3, n.11. [Online]. Available: http://www.firstmonday.org/issues/issue3_11/pang/

Singer, J. 1998. Online journalists: Foundations for research into their changing roles. Journal of Computer-Mediated Communication, v.4, n.1. [Online]. Available: http://jcmc.indiana.edu/vol4/issue1/singer.html