Hypertextual Dynamics in A Life Set for Two

Robert Kendall

Writing Program
New School for Social Research
66 West 12th St.
New York, NY 10011
Tel: 1-212-229-5880
E-mail: 102012.1273@compuserve.com


In most hypertexts the contents of nodes and the positions of links are fixed. Making these elements dynamic can help writers solve structural problems and help prevent navigational dilemmas for readers. The hypertext poem A Life Set for Two demonstrates several techniques for doing this. Floating links are positioned dynamically in response to the reader's progress. Variable nodes change their texts according to factors such as their context within the current reading. The texts of individual nodes are also influenced by global states--settings that can be changed manually by the reader or automatically by the program.


One of the promises implicit in hypertext is that freeing writing from the fixity of print will move it beyond the book to become a new, uniquely fluid medium. Yet in its typical manifestations, hypertext does not fully overcome the limitations of paper. The classic model of nodes and links may let the reader traverse different routes through the text, but the nodes themselves remain as fixed as pages of print. The reader's range of options for choosing a route also remains as predetermined as page numbering. The close relationship that still holds between this static hypertext paradigm and the paper book is demonstrated by the printed hypertextual works that conform to this model by providing instructions on how to jump from one distant page to another [Cortazar 66] [Packard 79] [Pavic 88] [Moulthrop 92]. Conversely, reading computer-based hypertext with a static structure is still partly a simulation of hopping among pages of print.

A number of hypertext systems have demonstrated that nodes and links need not be static. Dynamic hypertext, or what [Bernstein 91] calls "volatile" hypertext, lets the medium reduce its dependence on the methods and structures of print. A link need not be merely a fixed pointer to a new electronic page. Its location and destination can vary to suit the changing circumstances of reading. A node need not mimic an immutable page of print. Its text can change dynamically in response to the different situations in which the reader places it.

Unlike print, hypertext is conceptually a medium of multiplicity and contingency in which the potential elements of a reading are not shaped into a text until runtime. In this respect it is related to semi-improvised oral literature and, more distantly, to conversation, in which narrative elements or ideas are formed into coherent utterance in real time [Dickey 91] [Landow 92]. These modes of oral discourse depend for their success upon the speaker's awareness of what has already been said and a capacity to adapt and respond on the fly to changing circumstances. A hypertext may not be able to acquire skills of improvisation, but it can at least become dynamically responsive to the process of its own unfolding. This can make the hypertext reading experience smoother and more satisfying in many ways--for example, by better enabling the text to avoid presenting the reader with illogical sequences or unwanted recurrences of nodes.

On a more abstract level, the intrinsic fluidity of hypertext has been compared to that of thought itself. The earliest conceptual predecessor of hypertext, Bush's proposed memex, was conceived as an emulation of the "intricate web of trails carried by the cells of the brain" [Bush 45]. Bolter observes that "more effectively than the codex or the printed book, the computer reflects the mind as a web of verbal and visual elements in a conceptual space" [Bolter 91]. Joyce, in tracing the history of hypertext, maintains that "a remarkably consistent line of thinking holds that hypertext in some sense represents the workings of the human mind" [Joyce 95].

Artificial intelligence technology, let alone hypertext, cannot begin to offer a true replication of the intricate and dimly understood mechanisms of the human brain. Nonetheless, the parallels are interesting, and these are strengthened by dynamic hypertext techniques, which can offer a metaphor for the volatile nature of thoughts and memories. This points to intriguing conceptual directions for hypertext poets and fiction writers, whose business is trying to give readers a glimpse into the thoughts of others.

A Life Set for Two

The book-length hypertext poem A Life Set for Two [Kendall forthcoming] uses dynamic techniques in an effort to capture in writing something of the processes of thought rather than just its products. It does not purport to be a scientific simulation of neurological functions, nor does it bear any relationship to artificial intelligence--rather it uses hypertext as a literary device in an effort to convey more accurately the inner world of one human being to others. The dynamic techniques in this work extend well beyond the literary device, however, and could be applied to a wide variety of different kinds of hypertext.

The hypertext system behind A Life was written concurrently with the text of the poem in Microsoft Visual BASIC for Windows. Because the program is tied to the text at the code level, it could not serve as an authoring system for other works of hypertext, yet its features could be incorporated into a general-purpose hypertext authoring package.

A Life is a lyric poem in which the speaker reflects on his past. It opens with an essentially linear prologue, followed by a hypertext in which each node constitutes a memory or rumination. (This paper focuses only on the hypertext portion of the poem.) A highly nonlinear presentation of these fragments of thought is meant to parallel the random-access nature of human memory. The work is organized by a system of floating links, which makes it particularly easy to offer the reader a wide variety of different orders in which to read nodes.

Malloy has also explicitly exploited the relationship between nonlinear writing and the volatility of memory in her electronic fiction its name was Penelope. By randomly juxtaposing different fragments of remembered material, the work invites the reader "to step into [the protagonist's] mind, to see things as she sees them, to observe her memories come and go in a natural, non-sequential manner" [Malloy 1993].

Breaking down linearity only begins to approximate our perception of the mind's structure, though. Memories or ideas are never actually perceived as static phenomena. A Life employs variable nodes in which the text of the node itself changes depending on a number of parameters to parallel the way that memories and reflections seem to change under two circumstances: when they appear in different contexts of thought and when they occur in different frames of mind.

To help control the contents of nodes and the placement of links, A Life uses global states. At nearly any point, the hypertext can shift from one of these states to another. For example, some of the states represent the different attitudes with which the speaker in the poem reflects, expressing his marked ambivalence toward bygone experiences.

States: A Nation of the Mind

There are a few direct precursors in hypertext for the global states in A Life. The kiosk application the Electronic Multimedia Museum makes different information accessible through links depending on whether readers identify themselves as "children," "tourists," or "academics" [Shibata 93]. Moulthrop's hypertext fiction Chaos uses the concept of alternative versions of nodes, but the work is unfinished and the idea was never fully implemented [Moulthrop 90]. The prototype electronic encyclopedia designed by Weyer and Borning uses filters to determine whether measurements in the text should be displayed in metric or English units [Weyer 85]. Text-based computer games more commonly employ similar techniques, which have been categorized by Aarseth as "configurative user-functions" in his typology of textual modes [Aarseth 95].

The states in A Life function on two different levels. On the lowest level, three primary states represent different moods, or emotional perspectives, on the part of the poem's speaker. These moods are denoted by different colors rather than names, since they are too complex to be adequately summed up in a few words. Generally speaking, though, blue denotes relative serenity, red denotes tension, and black denotes coldness. One of these states is always active, and the reader can change from one to another at any time by means of a pop-up control bar, which contains three colored buttons, one for each mood. (See Figure 1.) To help keep the reader oriented and to heighten the work's aesthetic impact, each mood has a different-colored graphical background associated with it that appears behind the text. To emphasize the speaker's emotional ambivalence, the program randomly changes the mood if it has stayed constant for more than a set number of nodes.

These primary states also help spare the reader from textual repetitiveness. When all the links available at a given time lead to nodes that the reader has already read in the current mood, the mood automatically shifts to put forward versions of these nodes that have yet to be read.

Click here for Picture

Figure 1: A menu in A Life presenting floating links. Grayed-out options lead to nodes already read. The control at the upper right governs primary states (moods), while the control at the upper left governs secondary states.

On another level, A Life uses secondary states, which reflect thematic elements and can overlap with the primary states. One or another of the secondary states can be engaged or they can all be turned off. These states represent psychological undercurrents from the speaker's past that surface only occasionally in his thoughts. They pertain to the poem's central focus, a relationship with a woman. One secondary state invokes the theme of "other men," while another involves "potential parenthood."

If a secondary state is engaged by means of a pop-up control, parts of some nodes will change to reflect the relevant theme. There is also a node devoted entirely to the "other men" theme, and visiting it will engage the state associated with that theme. Secondary states become accessible only after the reader has traversed a certain percentage of the poem, to help the plot thicken as the reading progresses.

A Life demonstrates a few particular uses of global states, but the technique could find many other applications in both literature and nonfiction. Instead of different attitudes, the states could represent points of view--for example, events in a story narrated by different characters or arguments made in an essay from different sides of an issue. In reference works, different states could allow switching between summary or expanded versions of each discussion.

States could even become navigational aids serving much the same purpose as paths [Zellweger 89]. Sections of a hypertext with a parallel structure--such as three groups of nodes describing race relations in different cities--could be presented as one group of variable nodes controlled by states rather than as three separate paths. For example, the nodes could be Early History, Contemporary Problems, Efforts for Improvement, and so on. The states could be Los Angeles, New York, and Miami. This approach would not only simplify the hypertext structure that the reader has to negotiate but would also allow instant comparison of parallel sections of writing (such as contemporary problems in NYC and LA) merely by switching states.

Beyond the Page: Variable Nodes

A number of hypertext designers and researchers have recognized the value of variable elements within nodes and the technique has been put to some diverse uses. A simple early implementation is [Weyer 85]'s prototype electronic encyclopedia, which displays measurements in either metric or English units. Active Tioga documents provide a more complex example, filling variables with data from external sources [Terry 90]. [Zellweger 89] and [Garzotto 95] discuss the benefits of nodes that can modify themselves to accommodate different access paths or contexts, especially when multimedia elements are involved. Aarseth's textual typology includes the category "dynamic text" in which the individual textual building blocks of a nonlinear text (similar to nodes in a hypertext) are variable in their content [Aarseth 94] [Aarseth 95]. His examples of this category are text-based computer games, but his term "textons" could be applied to the changeable elements of variable nodes.

A related approach can be seen in hypertext programs that create virtual nodes on the fly by concatenating fragments from other sources, such as MacWeb [Nanard 93] and the World Wide Web's The Information Supercollider [Blackwell 95]. Also related is Rosenberg's use of changeable elements within the nodes of hypertext poems, though the results are quite different, since all the changeable portions of one node cannot be read together as a single text [Rosenberg J 93] [Rosenberg J 94a] [Rosenberg J 94b].

Manipulating the textual content of nodes in A Life involves two principal methods. The simplest merely provides alternative versions of an entire node, which are stored separately in the hypertext database. The current primary state, or mood, of the poem determines which version is displayed when the reader selects a link leading to the node. The different versions always share the same subject matter and imagery, and phrasing is often similar, but each presents its material from a different emotional bias.

The second method embeds variables within the text of the node. At runtime each variable is assigned one of several alternative text strings, depending on certain conditions. Thus, portions of the text within the node change under different circumstances, while the rest of the node's text stays constant.

Variables at the beginning of nodes hold transitional lines of poetry that change depending on the context in which the node is read--that is, what precedes it. These transitional variables allow a node to pick up a theme or image from the previous node. In this way, nodes become colored by what precedes them, somewhat in the manner that thoughts take on the tone or tenor of what they sprang from.

These small transitional variations are also an outgrowth of a natural proclivity whenever textual material is reorganized. When a writer reorganizes a story or an essay, or when a speaker reorganizes an old speech for a new audience, this will generally be accompanied by revisions to make the text flow better in its new ordering.

Transitional variables are especially important in A Life because of its sometimes nearly random-access structure. The nodes are all relatively self-contained so that they will work well in the greatest possible number of alternative orderings. Since none of them are organized into predetermined sequences or paths, the transitions are often necessary to make them follow naturally from one another, especially when they are highly contrasting in nature.

Here is an illustration of how transitional variables work in A Life. The nodes called "A Heart Well Done" and "Manna from the Stars" take this form when presented successively:

A Heart Well Done

"Just right," she said
when she sat down at the small, warm
      I'd set my heart on
like a dish on
         a wobbly
     table in a quiet corner.
She held it steady to explore
that delicately spiced place,
that moist flavor pulsing inside
    like a quivering
                  of breath
            against cold glass.
Her tongue worked slowly,
                     as if tracing initials
        on a steamed-up window.
Through all the layers of hunger's metaphors,
I meant what I tasted like
                           and she knew it.
Manna from the Stars

But what good were those refined tastes
when we sat down together
in our darkened life to watch
         The Once and Everafter
   All-Star Inner Movie of Ourselves?
As I passed her the popcorn,
I tried to cop a feel
      of her suspended disbelief.
We unfolded beyond
                   reach on the screen
and I joked about the bad acting,
the improbable script.
It was all I could do against the fear. . . .
The beginning of "Manna from the Stars" changes when it is preceded by a different node:

Dainties Under Glass

When I happened onto the right
back road after dark, I could see
the lighted window of her affection,
the shade carefully left
                         open a crack.
She knew I would work up the nerve
to peep in and
my plot thickening
between her caressing lips,
better times on the verge of coming.
Then the lights went off
         and the intimate blindness
     filled in the rest.
Manna from the Stars

I loved it when she joined me
in my voyeurism,
    sitting beside me in
our darkened life to watch
          The Once and Everafter
   All-Star Inner Movie of Ourselves.
As I passed her the popcorn,
I tried to cop a feel
      of her suspended disbelief. . . .
The text of the node then continues as above.

Variables are also embedded at the ends of many nodes in A Life, and these are governed by the current secondary state of the hypertext. If a secondary state is engaged, the relevant theme may be reflected by a few new lines appended to the node or changes made to the concluding lines already there.

Here is a node as it appears with no secondary state engaged:

I'll never forget the sadness
                             that wandered
           inside me, like a derelict
with nowhere to go,
                 then looked up
   into her eyes with their promise
of shelter for the night,
a good meal, an evening that received
all 57 prime-time channels of her body.
              "I'm open,"
                    she said.
        she was inches away from me,
the beds in her eyes were
When the secondary state that corresponds to the theme of "other men" is engaged, the final four lines assume this alternative form:

        she was inches away from me,
the image lounging in the beds of her eyes
   belonged to a richer man.
A third type of embedded variable used in A Life is independent of the state or context of the node. These variables, which appear only in some nodes, change their values every 1.5 seconds while the reader watches. The result is that the reader sometimes sees words or phrases within a line of poetry that cycle continuously among two or three alternative words or phrases while the rest of the node remains unchanged. Poets often use words or lines with ambiguous meanings to increase the resonance of a poem. These cyclic variables are largely an extension of this technique.

In the following example of a cyclic variable from A Life, the two words in braces are displayed alternately on the screen:

A candy wrapper,
    a rusty {can/can't} added
                     an artful touch
           of veracity to the scene
    as they lay there among
           her fallen fruit.
In theory, the technique of embedded variables could be replaced by simply using alternative versions of the entire node--that is, by creating a number of nodes that share the same constant text. In practice, though, this could become unmanageable if more than one variable or a few variations are involved. Being able to edit the variable and the constant parts of the nodes independently makes revision much easier for the author.

Variable nodes could have applications beyond those demonstrated in A Life or elsewhere. They could be particularly valuable in either literature or nonfiction to help ensure a coherent reading, regardless of the order in which important nodes are encountered. The technique could prevent the problem that arises when a reader comes upon a node that relies on key background information that is contained only in another node not yet read. Rather than isolating such background information in its own node, the system could provide it to the reader from within other nodes when needed.

For example, a narrative may contain half a dozen different nodes that could each potentially be a reader's first encounter with a particular character. If the reader has not been introduced to this character before arriving at one of these nodes, text held by a variable in the node could explain who the character is. If the reader is already familiar with the character, the explanatory text would simply be suppressed. In nonfiction, this approach could apply to the explanations of concepts or terms.

Variable nodes could also demonstrate causal connections among elements discussed in a hypertext. For example, in a hypertext story, if the reader selects a node that introduces a particular event, variable text in later nodes could be modified to reflect the consequences of that event. This could enable an author to create plot variations without resorting to many different narrative branches. In an essay, a reader could select among several nodes representing different assumptions, and the choice could trigger variations in succeeding nodes to show how it affects a line of reasoning.

Floating Links

Dynamic links, or what [DeRose 89] calls "intensional links," play a crucial role in an increasing number of hypertext systems. Microcosm [Fountain 90], Cybermap [Gloor 91], Max [Bieber 91], the Electronic Multimedia Museum, and the Textile Image Database System for Apparel Designers [Shibata 93] use dynamic links mostly to handle new content that may be added to or generated by a system. The Lynx link apprentice [Bernstein 91] and MacWeb [Nanard 91] [Nanard 93] can create temporary links between nodes containing related information. Trellis can monitor certain reader behaviors to determine whether to hide or display links [Stotts 91].

The "guard field" feature of Storyspace [Bolter 95] allows a different kind of dynamic linking, which implements some of the principles of "conditional paths" laid out in [Zellweger 1989]. Storyspace can create conditional links that are accessible to the reader only if certain criteria have been met, such as a specified node having been previously read or not read. This feature has proven valuable to hypertext fiction writers and has been used to good effect in Afternoon, a story [Joyce 90], Victory Garden [Moulthrop 91], "Lust" [Arnold 93], and "I Have Said Nothing" [Douglas 93] to enforce certain chronologies of events, vary the reader's route through the text, or help avoid repetition.

A Life uses a dynamic system of floating links, which bears some similarities to the systems above. In particular, it expands on many of the approaches that originated with guard fields, and like MacWeb it relies on categorizing nodes and links, but it also differs in significant ways from these and other systems. As with links in static hypertext, the end point of a floating link is fixed, always leading to one predetermined node. The source points, however, are not anchored to specific nodes. Rather, they are stored separately from the nodes and displayed as options in independent menus. The availability of a particular link at any given time depends less on the particular node currently on screen than on other parameters in the reading environment.

My creation of the floating link system for A Life was motivated primarily by two concerns. I wanted to give the reader maximum flexibility in changing the ordering of material in A Life, which required making a relatively large number of links available from each node. At the same time, I wanted to exert a high degree of control over the reader's progress to maintain direction in the work. Floating links let me achieve these seemingly irreconcilable goals by guiding the reader from one type of writing to another rather than from one individual node to another. They also let me employ a menu interface that ties in nicely with the central restaurant metaphor that pervades the poem. (See Figure 1.)

To give direction to the hypertext, I defined a large-scale reading template to which any potential reading of the text would conform. This template specifies a certain progression of elements within the poem, sketching out the broad strokes of the reading's structure but leaving the details open. Among these specifications are a gradual build-up in emotional complexity and a termination in a degree of closure after all or most of the material has been read. The template is implemented as a set of rules telling the system how to present floating links to the reader. One of the mechanisms it depends on is a monitoring routine that tracks which nodes have been read and what proportion of the entire hypertext they constitute.

To put the linking system in place, first I grouped the nodes into categories. I assigned the node names as options in pop-up menus, with a separate menu or submenu for each category. These menu options are the source points for each link. I then laid out rules that determine which categories will be accessible to the reader at different points in the hypertext, which in turn determines when each menu will be displayed. I made provisions for suppressing links on certain menus when necessary.

The two most important node categories in A Life distinguish between two different focuses: on the actions and motivations of the male speaker and on those of his ex-lover. The poem maintains a continuous interplay between these two categories, establishing a sort of metaphorical dialogue between the two characters. The menus for these categories are displayed alternately, making all the nodes in one category inaccessible from any other node in the same category and requiring the reader to alternate continually between the two when reading.

Highly charged categories--love, sex, and "other men"--are introduced one at a time only after the reader has read a certain percentage of the poem. This fosters a steady emotional build-up and is accomplished by hiding the appropriate submenus until the reader has traversed a specified number of nodes. Once these submenus appear, they are available from nearly any node. The only restriction is that the reader cannot make two selections in a row from these categories, which ensures that their elements are introduced gradually.

Floating links facilitate something that can be very difficult to create in hypertext: a true ending. A Life contains a concluding node that constitutes its own category. This becomes available only after the reader has consumed a predetermined amount of the poem, and from then on it can be selected at any time. Once the ending has been read, the reading is terminated. Since the ending is a variable node influenced by what precedes it and which states are current, it is effectively a set of alternative endings.

As the reader traverses the hypertext, menu options are grayed out to indicate nodes already visited in the current primary state. (See Figure 1.) This technique--familiar to users of the popular Netscape Navigator Web browser [Netscape 95]--steers the reader away from already familiar material, even though these links can still be selected if desired. This link marking technique, along with the menu structure, makes it virtually impossible for the reader to become disoriented or lost while reading the work.

A fringe benefit of the floating link approach was its facilitation of global revisions during the writing stage. If I wanted to make a new node accessible from many other nodes, I could just add it to the appropriate category rather than adding new links to each of these nodes.

Looking to the Future: Reading Templates

Floating links could shape the readings of many different kinds of hypertexts to various sorts of reading templates, while still allowing the reader great navigational freedom. For example, in fiction they could facilitate character or plot development by ensuring that noncontiguous nodes pertaining to a specific character or sequence of events are always encountered in a prescribed order and at a prescribed position (near the beginning, middle, or end) in a reading. They could ensure suspense in a narrative by introducing "teaser" nodes early on and delaying access to other nodes that resolve mysteries. They could contribute variety by allowing readers to return to certain passages only when they would appear in new contexts that shed new light on them.

This approach could also help avoid navigational problems by guiding the reader toward material yet to be read. It could help prevent two all-too-common misadventures for the reader: roaming through many already familiar nodes in search of new material, or leaving the hypertext without reading some key passages just because a few important links went undiscovered.

These goals are partially attainable through Storyspace's guard fields, and, as mentioned above, some Storyspace fiction has used this feature to help prioritize parts of the narrative. Floating links offer some distinct advantages over guard fields, though. Guard fields imposed on links can make a node inaccessible until another specified node has been read or they can prevent a reader from returning to a node already visited. They cannot, however, directly control whether a node is encountered near the beginning or end of a reading, and they cannot make a node accessible from anywhere in the hypertext. It is difficult to predict at what point in reading a large hypertext a reader will encounter a particular guarded link. The activation of floating links, on the other hand, is directly connected to the reader's overall progress though the text.

A hypertext need not be confined to a single reading template. The author could offer the reader several to choose from, each giving a different shape to the writing. Readers could also be allowed to customize these templates or design new ones themselves. The reading template of A Life allows some customization by letting the reader turn off the function that delays the appearance of some nodes--an option that might be appreciated by anyone going through the text for a second time.

Implementing floating links in A Life was fairly straightforward because the work contains a relatively small number of nodes. A very large hypertext, or one that contains many different narrative or expository threads, would probably best use floating links to supplement conventional static links. Links anchored to nodes in the usual manner, perhaps organized into paths, could serve as the reader's primary means of traversing the text. Floating links could be reserved for key navigational junctures.

Here is an example of how some of the scenarios described above might be implemented by combining floating links, static links, paths, and a means for monitoring the reader's progress:

A hypertext constructed with static links could be organized by assigning paths to sequences of nodes that constitute fairly discrete segments of narrative or expository material. The paths could be categorized by topic or theme and also designated as introductory, developmental, or conclusive in nature (that is, according to whether they should occur near the beginning, middle, or end of a reading) when appropriate. Priorities could be specified for certain paths if one should be traversed before another is made accessible.

Floating links could then be displayed at runtime when the system detects that the reader has reached a dead-end (that is, a node with no static links leading from it), has reread several nodes in a row (possibly looking for new material), or has reached a node that the author has designated as a point appropriate for jumping to another location in the text. These links could lead to the initial nodes of all the paths that contain entirely or mostly unread material and that fall into the same category as the path the reader is currently on. If everything in that category has been read, links to paths in a related category could appear. To avoid monotony, the system could occasionally display only links to material in different categories.

By monitoring how much of the hypertext has been read, the system could determine whether to make available links to introductory, developmental, or conclusive text. It could make sure that prioritized paths are made accessible in the prescribed order. If the reader has failed to encounter certain key paths after getting a good way into the text, the system could display links to those paths at the earliest opportunity.

Floating links could take on much the same function as a map or table of contents in helping readers find their way to unread nodes. Their operation would be much less obtrusive, though. They would function like a guide who shows travelers the way before they can become disoriented, rather than like a map that travelers resort to only after becoming confused about where to go next.

Reading a Moving Target

Variable nodes present some unusual challenges to the reader. To reread part of a hypertext, normally one can find the desired material with the aid of navigational tools or just backtrack to the relevant spot. In A Life the reader can easily return to any node, but unless it is encountered in the same context and state as on the initial reading, the text is likely to be different. I deliberately omitted a conventional backtrack feature, since the context-sensitive variables in the writing would make literal backtracking problematic.

I also believe that retracing one's steps back through the work would undermine one's sense of it as an unrepeatable real-time experience. Rosenberg has pointed out how the geometric, reversible nature of static hypertext is incompatible with the human experience of consciousness, which perceives time as irreversible and contingent [Rosenberg M 94]. The dynamic nature of A Life makes hypertextual reading into a process that the reader perceives as irreversible, since nodes generally change when they are revisited. This brings the work closer to the processes of thought that it is attempting to capture.

Rereading verbatim has its undeniable benefits, though. Poetry, especially, often requires several readings before all its layers of meaning are coaxed out. In A Life, literal rereading can also help clarify the interconnections between nodes that are created by the transitional variables. To retain the advantages of rereading verbatim, I created a separate Reread window. This contains not the work itself, but a transcript of the current reading. The reader can page back through the text to see each node exactly as it was presented, yet this has no effect on the current position in the poem when the Reread window is closed.

A Life also poses a challenge to anyone intent on reading every last word of the text. To uncover all the different variable text strings would entail reading each node in every possible context and state--a daunting process that even the most methodical of readers would be unlikely to attempt.

Instead of trying to devise a technical "solution" to this "problem," I ask the reader to accept this characteristic of the work as one that most distinguishes it from a printed book. The reader may do well not to try to fit an encounter with this poem into the mold of a conventional reading experience. Our earlier analogy with conversation might again be instructive here. Reading A Life is perhaps a little like engaging in a conversation (though admittedly a one-sided one) with someone you know will never reveal all of himself to you. In fact, it is partly the untapped material that makes the experience worth coming back to.

This brings us again to the notion of rereading. If one criterion for judging any book good is that you can always return to it and find "new things," then A Life does well on this score. Its "newness" is literal and not just a metaphor for what originally went unnoticed.

Textual Kinetics

To make reading A Life more conspicuously a real-time experience subject to the vagaries of temporality, all the text within each node is presented kinetically. It slides onto the screen one line at a time in a manner governed by the meaning and rhetorical weight of each line.

Another kinetic element is offered by an optional automated link-selection feature. If the reader engages this function, the program will wait a reader-determined length of time (from 1 to 60 seconds) for a link to be selected. If no selection is made, the program randomly chooses among the available links and displays the new node. This is similar to what [Zellweger 89] describes as "automatic playback control," though her approach involves following a predetermined path.

The reader can turn the link-selection feature on and sit back as if experiencing a live recitation, though the words are displayed rather than spoken. Or the reader can engage the feature and continue selecting links, creating an interesting conversational dynamic between reader and program. If the reader hesitates too long before making a selection, the program continues of its own accord, almost as if filling an awkward silence in a conversation.

Hypertext and electronic books are edging text ever further from the structural paradigms of print, but there is still a lot of open road ahead, especially in the direction of new poetry and fiction. As the medium comes into its own, dynamic techniques are bound to become more common. If we limit ourselves to static approaches, which are intrinsically the strategies of print, we may find that the structural and navigational obstacles sometimes encountered by complex hypertext writing will prove insurmountable.


I would like to thank Mark Bernstein for his invaluable guidance and advice. Thanks also to Espen Aarseth for his input. Partial financial support for A Life Set for Two was provided by the New Forms Regional Grant Program administered by the Painted Bride Art Center in Philadelphia.


[Aarseth 94] Aarseth, E.J. "Nonlinearity and Literary Theory" in Hyper/Text/Theory, G.P. Landow, ed., Johns Hopkins University Press, Baltimore, 1984, pp. 51-86.

[Aarseth 95] Aarseth, E.J. "Text, Hypertext, or Cybertext? A Typology of Textual Modes Using Correspondence Analysis" in Research in Humanities Computing 6, S. Hockey, N. Ide, and G. Perissinotto, eds., Oxford University Press, Oxford, England, forthcoming.

[Arnold 93] Arnold, M. "Lust" in The Eastgate Quarterly Review of Hypertext (Vol. 1, No. 2), software for Macintosh and Windows, 1993.

[Bernstein 91] Bernstein, M., Bolter, J.D., Joyce, M., and Mylonas, E. "Architectures for Volatile Hypertext" in Hypertext '91 Proceedings, ACM, Baltimore, 1991, pp. 243-260.

[Bieber 91] Bieber, M. "Issues in Modeling a 'Dynamic' Hypertext Interface for Non-Hypertext Systems" in Hypertext '91 Proceedings, ACM, Baltimore, 1991, pp. 203-217.

[Blackwell 95] Blackwell, T. The Information Supercollider, World Wide Web page, URL: http://www.eecs.harvard.edu/collider.html, 1995.

[Bolter 91] Bolter, J.D. Writing Space: The Computer, Hypertext, and the History of Writing, Lawrence Erlbaum Associates, Hillsdale, NJ, 1991.

[Bolter 95] Bolter, J.D., Joyce, M., Smith, J.B., and Bernstein, M. Storyspace, software for Macintosh and Windows, Eastgate Systems, Watertown, MA, 1991 (version 1.31, 1995).

[Bush 45] Bush, V. "As We May Think" in The Atlantic Monthly, July 1945, pp. 101-108.

[Cortazar 66] Cortazar, J. Hopscotch, G. Rabassa, trans., Avon, New York, 1966.

[DeRose 89] DeRose, S.J. "Expanding the Notion of Links" in Hypertext '89 Proceedings, ACM, New York, 1989, pp. 249-257.

[Dickey 91] Dickey, W. "Poem Descending a Staircase: Hypertext and the Simultaneity of Experience" in Hypermedia and Literary Studies, P. Delany and G.P. Landow, eds., MIT Press, Cambridge, MA, 1991, pp. 143-152.

[Douglas 93] Douglas, J.Y. "I Have Said Nothing" in The Eastgate Quarterly Review of Hypertext (Vol. 1, No. 2), software for Macintosh and Windows, 1993.

[Fountain 90] Fountain, A.M., Hall, W., Heath, I., and Davis, H.C. "MICROCOSM: An Open Model for Hypermedia with Dynamic Linking" in ECHT '90 Proceedings, Versailles, France, 1990, pp. 298-311.

[Garzotto 95] Garzotto, F., Mainetti, L., and Paolini, P. "Hypermedia Design, Analysis, and Evaluation Issues" in Communications of the ACM, August 1995, pp. 74-86.

[Gloor 91] Gloor, P. "CYBERMAP: Yet Another Way of Navigating in Hyperspace" in Hypertext '91 Proceedings, ACM, Baltimore, 1991, pp. 107-121.

[Joyce 90] Joyce, M. Afternoon, a story, software for Macintosh, Eastgate Systems, Watertown, MA, 1990.

[Joyce 95] Joyce, M. Of Two Minds: Hypertext Pedagogy and Poetics, University of Michigan Press, Ann Arbor, 1995.

[Kendall forthcoming] Kendall, R. A Life Set for Two, software for Windows, Eastgate Systems, Watertown, MA, forthcoming.

[Malloy 1993] Malloy, J. its name was Penelope, software for Macintosh and Windows, Eastgate Systems, Watertown, MA, 1993.

[Moulthrop 90] Moulthrop, S. Chaos, software for Macintosh, unpublished, 1990.

[Moulthrop 91] Moulthrop, S. Victory Garden, software for Macintosh computers, Eastgate Systems, Watertown, MA, 1991.

[Moulthrop 92] Moulthrop, S. "Informand and Rhetoric: A Hypertextual Experiment" in Writing on the Edge, Fall 1992, pp. 103-127.

[Nanard 91] Nanard, J., and Nanard, M. "Using Structured Types to Incorporate Knowledge in Hypertext" in Hypertext '91 Proceedings, ACM, Baltimore, 1991, pp. 329-343.

[Nanard 93] Nanard, J., and Nanard, M. "Should Anchors Be Typed Too? An Experiment with MacWeb" in Hypertext '93 Proceedings, ACM, Seattle, 1993, pp. 51-62.

[Netscape 95] Netscape Communications Corp., Netscape Navigator, Web browsing software, 1995.

[Packard 79] Packard, E., originator. Choose Your Own Adventure series, Bantam Books, New York, 1979-89.

[Pavic 88] Pavic, M. Dictionary of the Khazars: A Lexicon Novel in 100,000 Words, Christina Pribicevic-Zoric, trans., Vintage Books, New York, 1988.

[Rosenberg J 93] Rosenberg, J. Intergrams in The Eastgate Quarterly Review of Hypertext (Vol. 1, No. 1), software for Macintosh, 1993.

[Rosenberg J 94a] Rosenberg, J. Diffractions through: Thirst weep ransack (frailty) veer tide elegy, software for Macintosh, forthcoming.

[Rosenberg J 94b] Rosenberg, J. "Navigating Nowhere/Hypertext Infrawhere" in ECHT '94 Proceedings, ACM, Edinburgh, Scotland, 1994.

[Rosenberg M 94] Rosenberg, M.E. "Physics and Hypertext: Liberation and Complicity in Art and Pedagogy" in Hyper/Text/Theory, G.P. Landow, ed., Johns Hopkins University Press, Baltimore, 1984, pp. 268-298.

[Shibata 93] Shibata, Y., and Katsumoto, M. "Dynamic Hypertext and Knowledge Agent Systems for Multimedia Information Networks" in Hypertext '93 Proceedings, ACM, Seattle, 1993, pp. 82-93.

[Stotts 91] Stotts, P.D., and Furuta, R. "Dynamic Adaptation of Hypertext Structure" in Hypertext '91 Proceedings, ACM, Baltimore, 1991, pp. 219-231.

[Terry 90] Terry, D.B., Baker, D.G. "Active Tioga Documents: An Exploration of Two Paradigms" in Electronic Publishing, May 1990, pp. 105-122.

[Weyer 85] Weyer, S.A., and Borning, A.H. "A Prototype Electronic Encyclopedia" in ACM Transactions on Office Information Systems, January 1985, pp. 63-88.

[Zellweger 1989] Zellweger, P.T. "Scripted Documents: A Hypermedia Path Mechanism" in Hypertext '89 Proceedings, ACM, New York, 1989, pp. 1-14.