But I Know What I Like
by Robert Kendall
(also in SIGWEB Newsletter, Vol. 8, No. 2, June 1999)
Readers of hypertext become authors, creating the work as they go. This old saw of hypertext criticism suggests an aesthetic relativism with little room for notions of absolute quality. How can a work be judged good or bad if it is "created" by its reader? We may not accept this quandary as a corollary of postmodern theory--we may dismiss the theory itself. But even if we believe there can be criteria for excellence in hypertext, another question arises. Is it wise to try to foist rules on an art form that seems brought into the world at least partly for the purpose of breaking established literary rules?
One thing is certain: These concerns won't stop critics from applying their own aesthetic expectations to hypertext, regardless of how appropriate those expectations may be. If we want hypertext to be judged on its own terms, we must establish what those terms are. So at the risk of being branded a counterrevolutionary, I've attempted here to pin down what constitutes good hypertext, apart from the usual prerequisites of good writing in general.
The hallmark of worthwhile hypertext that everyone is likely to agree upon is that it offers a reading experience fundamentally different from reading print. Hypertextuality should not be mere decoration that could be easily stripped away to reveal a would-be piece of printed linear text beneath. High-quality writing alone doesn't guarantee a compelling hypertext, just as simply placing well-written text in the mouths of actors doesn't guarantee compelling drama. An author must understand the unique demands and potential of the genre.
Hypertext literature has two fundamental properties that make it unique. First, it lets readers interact with the text. Second, it lets them perceive the text as a nonlinear or multilinear structure. The quality of the interaction and the effectiveness of the nonlinear structure should therefore command a critic's attention. Let's examine the components of these properties to help us hone our critical apparatus.
In trying to define satisfying interaction, we have three main issues to consider: agency, which lets the reader influence or mold the final form of the text; momentum, which propels the reader onward; and closure, which provides a sense of having satisfactorily ended the reading.
Janet Murray describes agency in electronic literature as "the satisfying power to take meaningful action and see the results of our decisions and choices."  There are two components to agency. First is the degree to which readers' actions can change the work or shape their perceptions of it. Second is the awareness readers have of the effects of these actions.
Obviously the number of different possible readings contained within a hypertext helps determine the degree of agency that it will allow the reader. The sheer quantity of combinatorial possibilities presented by links and nodes isn't the only consideration, though. A collection of nodes that are all self-contained and independent may lend itself to the highest possible number of reorderings, but if there aren't numerous relationships and implicit connections among these discrete text components, the order may have little effect on the reader's perception of the content. Any printed collection of independent poems will allow the reader to jump randomly from poem to poem and consume the work in a completely nonlinear fashion--and, in fact, this is how such collections are often read. The degree to which each such reordering actually alters the impact of the overall work, however, may be minimal.
In contrast, a hypertext may offer the reader relatively few alternative paths through a work, but if each offers a radically different perspective on the piece, the result is a high level of reader agency. The stronger the intrinsic narrative or thematic connections among nodes, the greater the significance when these connections are realigned by different sequencings. Multiple linearities that are intertwined in complex webs--relationships that extend through many nodes--can offer the richest interactive experiences. When the context of one node changes within such a structure, the import of that change can resonate through many other nodes.  Complicating linearity in a hypertext can be much more fruitful than trying to eliminate it.
Readers must be able to get a handle on all this complexity. A work may bombard them with navigational options and opportunities for variant readings, but if they have no clear sense of what the options represent and how the effects of one choice differ from those of another, they will feel powerless. Too many vaguely defined choices can seem like no choice at all--or rather, just the single choice of following more or less randomly wherever the software leads.
When words within the text serve as link anchors, the suggestiveness of these words directly bears upon the reader's sense of agency. Proper names and thematically important words are likely to conjure up a destination in the reader's mind more effectively than neutral words. Supplementing a link with some descriptive hint of its destination--which could appear in the browser's status bar or elsewhere--also makes for more informed reader choices. Particularly empowering for the reader are links defined by navigational function. For example, a menu at the bottom of the screen can explicitly offer options to continue with the current thread, digress to one of several new paths, or return to some centralized location. Storyspace hypertexts often let the reader either hit the Enter/Return key to continue a thread or select a link within the text to initiate a digression. Graphical maps routinely increase agency in hypertext, offering a broader context for possible destinations by visually depicting relationships among them.
Once the reader has followed a link, a clear relationship between source and destination will increase the reader's feeling of control. If, on the other hand, it's not apparent why the choice led to the destination, agency will suffer. In the case of a truly powerful link, the destination text will also shed new light upon the significance of the linked word within the source node.
Of course, the relationships between choice and outcome should not always be simplistically obvious and predictable in a hypertext. There is always room for subtlety and surprise. There is a difference, however, between navigational ambiguity that adds depth of meaning and link anchors that are forced onto words that aren't really appropriate for them. Sometimes a structural concern will demand a link to a particular node, even though no words within the text are really suitable to function as the link. The author may then just slap the link on whatever word is the closest fit--an artistic compromise that can have the same awkward effect as a forced rhyme in traditional verse.
The importance of clear and meaningful choice doesn't mean that randomization is necessarily detrimental to hypertext. If each mouse click randomly calls up a new section of text and readers fully understand the process, they can sit back and appreciate the fortuity of the resulting juxtapositions. Though readers may not be able to directly influence the ordering of the text, they can clearly see how their actions affect it. This is what can make such randomly ordered works as Malloy's its name was Penelope satisfying to read. 
Keeping It Moving
The next ingredient in the interactive mix is momentum, which keeps the reading going. Momentum depends upon the reader being both motivated to move forward and able to do so. It might seem that agency and momentum are inherently at odds with one another. After all, the foolproof way to maintain reading momentum is to keep readers turning the pages of a linear text in which each passage is carefully calculated to draw them into the one that follows. Especially in fiction, the more readers can alter the overall structure, the more opportunity there would seem to be for problematic or confusing narrative configurations that don't seem to go anywhere.
In fact, though, since true agency depends upon choices being meaningful and powerful, it aids momentum. The reader may have a dozen links to select from, but if none leads to what is sought--whether it be the continuation of a thread, a map, or simply unread material--both agency and momentum are thwarted. Agency and momentum both depend upon an author successfully anticipating the reader's desires. This usually means making available a number of different types of options that can respond to differing situations.
Just as in printed text, momentum doesn't rely on merely arousing and then satisfying an urge to find out what happens next. Maintaining a satisfying balance between continuity and variety in the unfolding text is ultimately what keeps a reader hooked. A work that always yields exactly what is expected becomes predictable and monotonous, while too many surprises and non sequiturs can lead to confusion or even incomprehensibility. While hypertext is at a disadvantage in some ways when it comes to making the reading flow, it does have a leg up on its printed counterpart in one respect: Whenever readers feel the need for more variety in a hypertext, they can simply jump to new episodes or vary their navigation strategies. In some instances, readers may also be able to augment the degree of continuity in a reading, though this is more difficult. The better a hypertext is able to satisfy readers' changeable needs for the smooth or the rough, the better it will hold their interest. Hence another advantage of function-defined links that signal continuation or digression.
It is all too common for a hypertext reading to stall because the text starts to loop and successive choices yield already familiar material. This can make readers feel as if they are running in place. To minimize text looping, an author can avoid constructions in which a disproportionately large number of paths lead to a single destination with relatively few links leading away from it. Providing frequent exit points from all long paths can help, too.
Momentum need not always depend upon moving through new text. A reader can feel a sense of progress when a deeper understanding of the work evolves through reexamining familiar material. A reader may backtrack to explore subtle relationships between linked text components, and links that lead to old nodes can reveal new structural joints. A hypertext should lend itself to a certain amount of fruitful rereading, since textual reencounters are necessary for gaining any real sense of the multiple configurations contained within the work. Ambiguous passages that pick up different implications from different surroundings add impetus by making rereading a process of discovery. Variable nodes, in which the text itself changes when the node is reread, can provide further motivation for rereading. 
The reader's attention will periodically have to shift from the text to the mechanics of navigation. This shift can either distract readers from the work or draw them more deeply into it, depending upon how smoothly the navigational tools function and how seamlessly they are integrated into the overall artistic conception. For example, suppose the reader frequently returns to a map screen in the hypertext. Suppose the map is easily accessible and contains richly suggestive imagery in which the reader often notices something new when returning to study it. The visits to the map then become a rewarding part of the reading itself. If, on the other hand, the map is hard to find, dull to look at, and confusing, then the reader is likely to experience its use as a disruption. Even when simply experimenting with features of a well-designed hypertext interface to learn more about their workings, readers may feel their knowledge of the piece advancing.
All Things Must End
The momentum of a reading must inevitably dissipate at some point so the reader can stop. Ideally, this should happen in a way that provides some feeling of closure. The strongest motivation for stopping is the reader's belief that all of the text has been read. The more secure this belief, the stronger the sense of closure. Closure also depends upon how effectively the last-read passage serves as a logical stopping point, if not necessarily a conclusion. Important too can be a reader's sense of having "satisfied the curiosities, interests, predictions and quests" that spurred on the reading.  The author may want to avoid wrapping things up neatly or providing a true ending, but even open-endedness is better served when the reader feels that the work is over rather than merely abandoned. The feeling of not having been able to finish reading something is quite different from the feeling that there is no true end to it.
It can be hard to gauge how much of a hypertext remains unread. Unless there's a feature that specifically indicates this or the work is very short, the reader can only assume that all or nearly all has been consumed when traversing a large number of nodes yields only previously read material. The more widely a reader ranges through the hypertext encountering little or no new material, the stronger will be the belief that virtually all has been read, and hence the stronger the feeling of closure. Unfortunately, if this final process is prolonged by occasional discoveries of unread text here and there, it can seem more like a tedious janitorial mopping up than a satisfying culmination of the reading experience. It can impart a frustrating perception of the work merely fizzling out, paradoxically undermining the sense of closure gained from knowing one has read everything.
A carefully balanced link structure can facilitate closure. Providing many ways to get to all the important sections will minimize the chances that small pockets of important text will elude the reader toward the end. An easily accessible central map or table of contents can help by providing entry to paths that traverse all the most significant textual territory. Conditional links can also serve to direct readers to unread material, as in Arnold's Lust.  Ideally, when the text starts to loop relentlessly, it should be a good indication that there is no more to read.
A visual indicator of how much text remains can be a godsend near the end of a reading. Strickland's The Ballad of Sand and Harry Soot provides a particularly unobtrusive but effective example of this approach.  A frame at the bottom of the screen contains a string of zeros, one for each node in the work. Each zero is linked to its corresponding node, and the link colors provide a visual gauge of how much has been read.
The Big Picture
Now let's turn our attention from the nitty-gritty of interactivity to issues of overall structural integrity. The effectiveness of just about any artwork depends partly upon the successful integration of form and content, and hypertext is no exception. In fact, if a work's hypertextuality isn't an integral part of its conception, it may seem at best just mere decoration and at worst an extraneous annoyance.
It is common in hypertext to find central metaphors in the writing that reflect properties inherent in the medium. For example, fragmentation is a major theme in Jackson's Patchwork Girl  and Cramer's In Small & Large Pieces.  Bly's We Descend examines historical "truth" as a collage of sources and interpretations.  Other works employ motifs of exploration or journeying that play off the navigation factor.
Of course, the tropes need not be so obvious to be effective. Indeed, if a work is too blatantly "about" hypertext, it risks coming across as an exercise in theory. All sorts of subtle multiplicity, multivalence, and synchronism lend themselves well to embodiment in hypertext. The best links are realizations of connections that already exist in the text. If associations and relationships on many different levels unify diverse parts of a work, hypertext is a natural outgrowth of the writing. It's very satisfying to find that one's evolving understanding of a work's hyperstructure sheds light on its thematic underpinnings and vice versa. This imbues the navigational process with a deeper significance.
A rich, complex thematic texture benefits a hypertext in other ways. Unlike the connections that constitute plot line, character development, and some other standard narrative elements, thematic and metaphorical ties rarely depend upon chronology. Thus thematic threads can help hold a work together despite many disjunctions of narrative time and setting. Thematic density can also play an important role in making a virtue of the inevitable textual recurrences within a hypertext. Nodes that resurface in the midst of new themes are likely to come across as development rather than mere repetition. 
The more comprehensible the large-scale hypertext structure is to the reader, the more meaningful it will become. Visual representations of hyperstructural patterns can help here. A short hypertext can often be mapped to a visual pattern in which every node finds a representation. If the geometric logic of the graphic reflects something of the text's thematic or narrative logic--showing, for example, oppositions or parallels--this can be very revealing. Works by Deena Larsen, such as "Stained Word Window," are a good example of this. 
Attempts at visually representing every structural detail in a large, complex hypertext generally become too convoluted to be easily grasped. Also, a work's structure may lend itself to many different geometric renditions, and pinning it down to one visual representation may actually obscure the full richness of relationships. A simplified, perhaps somewhat ambiguous rendering of relationships among major structural components is often the most effective approach. Yet not all hypertexts will benefit from a map and the simplifications it may demand. Some forms of hypertext defy mapping altogether, such as an "intratext" in which the reader changes words or phrases within a single screenful of text. 
Other aids can also help the reader conceptualize the structure. Recurring visual elements in nodes--such as color, font, graphics, or even patterns of link placement--can signal long-range structural relationships among widely dispersed components. If paths are a significant part of the work, readers will better understand their function if each has a meaningful name and they always know which path they are currently pursuing and which paths they have the option of following at any given point.
As we grapple critically with these different elements in hypertext, we must recognize the natural give and take among them. Some types of particularly engaging and potent interaction come only at the expense of an easily discernible overall form. The slight disorientation induced by a dearth of navigational signposts can sometimes perfectly mirror a work's contents. Just as slow-paced action or the lack of a clever plot doesn't necessarily weaken a printed novel, the absence of some usually desirable quality doesn't necessarily make for a lesser hypertext. A good writer in any genre knows how to make trade-offs. I merely hope the critical criteria outlined here can provide some useful guidelines--even if only for writers looking for new rules to break.
 For discussion of how contextual changes can resonate through large sections of a work, see Robert Kendall, "Parsing the Cold: McLaughlin's Notes Toward Absolute Zero," in SIGWEB Newsletter (Vol. 7, No. 3, 1998), pp. 7-9. [back]
 For analysis of an excellent thematic structure in a hypertext fiction, see Kendall, "Parsing the Cold: McLaughlin's Notes Toward Absolute Zero." [back]
copyright (c) 1999 by Robert Kendall