Card Shark and Thespis

Card Shark and Thespis


Eastgate Systems alumns Diane Greco and Mark Bernstein explain two “exotic tools for hypertext narrative.”

Hypertext Fiction and Its Critics

Although games, visual art, and textual experiments had long been areas of academic research, the first artistically convincing explorations of literary computing appeared in the late 1980s. It was only in these years that computers became sufficiently commonplace that a computational creation could realistically hope to find an audience. Of equal importance was the gradual acceptance of Ted Nelson’s thesis (Nelson 1976) that computers could be tools for artistic expression, for even in 1982 the title of Nelson’s Literary Machines was meant to shock and surprise.

The final and critical step, first taken by an informal workshop of American writers who called themselves TINAC, was to recognize that hypertext links need not be merely annotations (as in Yankelovich, Meyrowitz, and van Dam 1985) or plot points (as in Adventure). Links, they realized, could serve as exquisite literary connections, explicitly opening the text to the readerly interactions and interventions that are explicitly (albeit tacitly) part of all serious reading. Links could change point of view, enact a time shift, or hold contradictory elements in suspension. Links could suggest new formalisms, new structure, a new large-scale punctuation. Indeed, even the absence of an expected link, or the readerly effort require to decode the gap between the point of departure and the point of arrival, could prove as eloquent as a dramatic musical rest (Bolter and Joyce 1987; Joyce 1988; Harpold 1991).

Although the TINAC group agreed on these principles, they differed strikingly in execution, giving rise to three quite distinct approaches to literary hypertext that continue to shape the literature today. Michael Joyce (1990), in afternoon, built dense, lyrical explorations of reality and memory. Stuart Moulthrop (1991), in Victory Garden, uses links in an ironic, less purely evocative mode; Joyce speaks of links as “words that yield,” but Moulthrop’s witty (and, often, bitterly sarcastic) links yield nothing to anyone. Where Joyce’s work is overtly metafictional, Moulthrop and J. Yellowlees Douglas (1993; “I Have Said Nothing”) use links as hyperbaton (Bolter 1997). That is, where in rhetoric we might depart from idiomatic word order to achieve a dramatic effect, so in hypertext we may contrive to reveal an underlying narrative that gradually emerges as the reader interprets scenes traversed in a sequence outside the writer’s immediate control. Finally, in Uncle Buddy’s Phantom Funhouse, John McDaid (1992) devises an artifactual hypertext, literally a literary machine, simulated on the reader’s computer, which the reader must learn to operate and decode.

All three approaches received substantial critical applause, a lasting following, and (perhaps most importantly) have inspired numbers of subsequent hypertext artists. Joyce’s lyrical hypertextuality finds recent echoes, for example, in Chapman’s (2001) Turning In, Strickland’s (1998) True North, as well as Arnold and Derby’s (1999) Kokura. Moulthrop’s hyperbaton is key to Coverly’s (2000) Califia, Cramer’s (1993) “In Small & Large Pieces,” Eisen’s (2001) “What Fits,” and Amerika’s (1997) Grammatron. McDaid’s artifactual approach, dormant for some years, finds recent expression in Bly’s We Descend, Malloy and Marshall’s (1996) Forward Anywhere, and in Sondheim’s executable poems. Many combine several approaches, as when Shelley Jackson’s (1996) Patchwork Girl masterfully shifts among all three.

A substantial critical literature has grown up around these works, as scholars and critics have sought to understand and appreciate them more fully. Robert Coover’s early essay (Coover 1992) and his comprehensive, thoughtful review of the entire body of early hypertext fiction (Coover 1993), helped situate hypertext fiction near the center of the contemporary literary landscape. Bolter’s (1991) Writing Space situated hypermedia in the continuing development of writing, and Landow’s (1992) Hypertext: The Convergence of Contemporary Critical Theory and Technology called attention to the theoretical complexity of reading and criticism. Lanham (1993), Gaggi (1997), Murray (1997), Douglas (2000; in The End of Books - or Books without End?, a revision of her 1992 dissertation), and Aarseth (1997) each contributed monographs with important early readings of key hypertexts.

Despite the ahistorical claim that early hypertext critics were blindly supportive (Eskelinen and Koskimaa 2000), hypertext fiction has always attracted stern criticism as well as praise. Much of the early critique centered on the apparent similarity of the computer to the television, expressing the fear that images, mixed media, and luminosity will of necessity subvert literary values (Birkerts 1994). Others, perceiving the relationship between hypertext and postmodernism and finding postmodern theory distasteful, sought to ridicule postmodern theory by denouncing hypertext (Miller 1998). Inevitably, individual works received both praise and blame, and it is not uncommon to read of critics who, after the passage of months or years, found great merit in work they had previously disliked (Walker 1999).

Narrative and the Nature of Hypermedia

An important and common perception among literary critics with scant experience of hypertext is that hypertext is necessarily incoherent, and that its incoherence is due to its technological substrate. They assume that, although even nontraditional print narratives convincingly and pleasurably mimic the apparent linearity of temporal experience, the technological qualities of hypermedia limit it to a particularly intolerable nonlinearity in which narrative is necessarily fractured, unsatisfying, and unpleasurable.

Whereas recalling the obvious fact that print is itself a technology with a long history of change and development would be enough to refute this flimsy argument, there is also much positive evidence that satisfying narratives are indeed produced in hypermedia. For instance, one may point to the continued popularity of early hypertext fictions (e.g., afternoon, Victory Garden), to the appearance of new fictions (e.g., Turning In, Holeton’s (2001) Figurski At Findhorn on Acid), to the myriad university courses in hypertext fiction, writing and criticism, or to the flourishing secondary literature on the subject (such as that by Nelson award winners Walker (1999), Tosca (2000), and Miles (2001)). Moreover, the importance of narrative in the craft of hypertext writing has been recognized from the beginning (Bolter and Joyce 1987), for narrative is central not only to works of imagination but also to myriad other forms of human expression, including technical (Bernstein 1991) and scholarly writing (Kolb 1997).

Hypertexts are not structureless, and to call them “nonlinear” is too general to be informative. After fifteen years of hypertext publishing, it is clear that once writers reject the idea that the linear narrative typical of print is the only acceptable model for storytelling, there is no theoretical limit on the varieties of structure that an author may develop. Nonetheless, we require a working vocabulary. In “Patterns of Hypertext,” Bernstein (1998b) catalogs some common patterns in observed hypermedia narratives and argues that patterning and narrative coherence are tightly coupled when dealing (as one often does in hypermedia) with stories that unfold according to organizations of composites and aggregates. Just as tapping a glassful of supersaturated solution can precipitate entire crystals, traversing a link can reorder events in a whole narrative. If a narrative contains more than one link (and most hypertexts do), ordering events is a complex task indeed. Patterns allow coherence to emerge when a narrative supports many different possible orderings of events. Unless otherwise noted, the words “structure” and “pattern” are used interchangeably in this paper to refer to ways in which hypermedia narratives can be said to cohere (though see Alexander, Ishikawa, et al.). For detailed discussion see Mark Bernstein, “Patterns of Hypertext” and “Hypertext Gardens.”

Do the patterns we observe in hypertext fiction arise directly from hypertextuality, or are they artifacts arising from the mediation of the system?

The nature of hypertext is best discerned by studying actual hypertexts, which at this early point in the history of hypermedia, cannot yet be separated from the tools of their production, presentation, and increasingly, distribution. Over the past decade, the descendants of three systems – HyperCard, Storyspace, and Mosaic – have been the tools most frequently selected by hypertext fiction writers. The dominance of these particular systems among fiction writers need not be ascribed to any inherent virtue or suitability to the task. Accessibility plays a crucial role, as do the accidents of history. If Guide, Trellis, or NoteCards had survived to develop a literary following, our current impressions of the nature of hypertext narrative might be quite different. Different writers use the same system in drastically different ways, Compare, for example, Joyce’s afternoon, Arnold’s Lust, Landow and Lanestedt’s The In Memoriam Web, and Strickland’s True North. All were written with Storyspace, but their use of links varies tremendously. but systems inevitably shape hypertexts. Are the properties of hypertext fiction, such as those observed in Bernstein’s “Patterns of Hypertext,” and deplored by Miller (1998) in “,” intrinsic to hypertext, or do they arise from the idiosyncrasies of specific systems?

This paper reviews two exotic hypertext systems, tools suitable for hypertext narrative but dramatically unlike the tools currently in use. Our motivation for describing these tools is also unusual. The customary reason for building a new system is to build a better system; here, we wanted to build a strange system, a hypertext environment that might let us step back from Storyspace and the web in order to gain a better perspective. We do not wish to argue that these systems are better than, say, Storyspace: Storyspace is simpler, more elegant, more flexible, more widely available. We do not suggest that hypertexts written with the new system will be better than those written with other tools. For our purposes, we need not be better, we need only be different.

In the remainder of this paper, we first explore a language or notation, Card Shark, that describes sculptural hypertexts. Shark is small, simple, and appears not to be very expressive, but it can readily describe complex hypertext structures. Next, we embed Shark in a dramatic context: we create a simple theatrical environment that represents characters moving through space, a space through which the reader moves to witness and perhaps to participate in the action. The nature of this participation, though superficially similar to interactive fiction, may avoid internal contradictions that confront conventional immersive fictions. Finally, we conclude with some thoughts on how this approach might be evaluated.

Sculptural Hypertext and Card Shark

Conventional hypertexts take a set of unconnected nodes (or pages, or lexia) and link them together. Card Shark begins with a set of lexia, all of which are connected to each other, and builds structure by removing unwanted connections. We call this initial set a tangle. We call Card Shark sculptural because we create structure by removing unwanted connections, much as a sculptor may create form by removing unwanted stone. Traditional hypertext tools, in this sense, are calligraphic; we create structure by adding lines until we have added exactly the necessary degree of connection.

Where a sculptural strategy has been employed in the past – most notably, perhaps, in Malloy’s (1993) Its name was Penelope and in Malloy and Marshall’s (1996) Forward Anywhere – it has been chosen in part to deemphasize temporal sequence and narrative structure (Golovchinsky and Marshall 2000). Card Shark, as we shall see, foregrounds sequence and emphasizes structure.

A Card Shark node (or card) contains some text, typically a brief, focused passage. Each card may also specify constraints on the context in which it may appear. For example, AFTER 10 requires that the node may be visited only after ten other nodes have been seen. A node that appears BEFORE 25 may only be visited early in the reading; if it is not seen early, it will not be seen at all. A variety of constraints may be applied to a node; as in a conventional hypertext, it is likely that some nodes will never appear in any given reading.

Each card may also specify modifications it makes to the reading context, chiefly by posting assertions on a blackboard. A passage that serves to introduce a new character, for example, could ASSERT WENDY. Other cards that REQUIRE WENDY can be visited only after this introduction. Later, a passage may remove WENDY from the scene and RETRACT WENDY.

Given a collection of cards, we read them by following a simple set of rules The details of these performance rules are often arbitrary and the reader may easily envision alternatives. These procedural details affect pacing and rhythm during performance, but effective writing is far more important to the performance than the details of these performance rules.:

1. The collection of cards is shuffled and the blackboard is wiped clean.

2. The reader receives seven cards from the deck.

3. The constraints for each of the player’s cards are evaluated. Cards whose conditions are not satisfied are disabled; the reader sees at most a brief title and an indication of what conditions need to be satisfied for the card to be seen.

4. From among the cards whose constraints are met, the player selects a node to visit next.

5. The selected node is visited. Its full text appears (or is performed) on the screen. If the node makes assertions or modifies the environment, those actions are performed. The card remains “on the table”; we may look at it again whenever we like, but it will never be “played” again.

6. The player receives a new card, and repeats until the reading is over. If the granularity of the lexia – the size of the “card” – is large, Card Shark ‘s constraints describe the episodic architecture of the narrative. If the granularity is very small – individual words or phrases – the constraints describe a text generation engine. If the lexia were lines of iambic pentameter, the constraints could describe a rhyme scheme.

We can easily envision other variations. In particular, we might use one deck but maintain two separate blackboards (perhaps called Plot and Subplot). The two blackboards provide separate contexts in which cards could appear, offering the reader and the writer greater flexibility.

In conventional hypertext tools, connecting nodes in a sequence is easy but connecting nodes in a dense tangle requires effort. Indeed, despite a priori concerns that hypertext disorientation would lead to confusion, incoherence, or inattention, the variety of expedients authors adopt to deliberately disrupt the reading line clearly suggests that disorientation is hard to achieve (Landow 1990; Bernstein 1991). In Grammatron, we see time stress and fluid links (Zellweger et al. 2001); in Guyer’s (1992) Quibbling we observe dense links that refuse to signify their intent; in Rosenberg’s Intergrams, lexia are superimposed and links are dynamically entwined. If disorientation followed naturally from the nature of hypertext, would such expedients be necessary or useful?

Card Shark inverts the situation; making a tangle is easy, but making a strict sequence takes work. The tangle, not the link, is a Card Shark primitive.

Liveness, Card Shark, and Transitions

When writing for Card Shark, we are naturally concerned with liveness – with avoiding premature termination or inadvertent closure (Douglas 2000). Consider a reader in the midst of reading a Card Shark hypertext. A time may come when the reader examines her seven options and finds that none of the preconditions are met. The position is dead, the reader is stuck. The situation is directly analogous to reaching a conventional hypertext node with no outbound links, or to an short, inescapable cycle that signals closure [Bernstein 1998b]. Early hypertext writers worried about deadness almost as much as they worried about disorientation [Bernstein 1991]; today, we press the Back button and wonder what the fuss was about. Sooner or later, this is inevitable: we will run out of cards and the story must eventually end. But the story should have a chance to play out first; we must take care to let the story begin before it ends, to avoid stranding the reader at the start.

Imagine, for example, a Card Shark hypertext that describes a twilight encounter in the garden on Tuesday night, and its dénouement in the nearby bedroom the following morning. Some actions REQUIRE NIGHT; others REQUIRE MORNING. If our current context is the night and our available actions require that it be morning, we need a transition that moves from night to morning. Conversely, if we have been reading about events in the dawn-lit bedroom and there is more to learn about last night’s unexpected encounter in the garden, we require a transition that moves from the morning to the previous night, from the bedroom to the garden. Indeed, if readers are not to constantly encounter dead positions, we need to provide a rich assortment of transitions to facilitate movement, to shift between times, and to get characters on and off stage.

Multivalence is not a vice, of course (Bernstein 2000). That is, these transitions need not limit themselves to their immediate business. It is important to observe that the text of a Shark node need not describe the change it accomplishes. The text might, for example, proceed from the consequence of the asserted change without describing the transition. Some transitions may not need to be expressed, either because the reader will understand them or because we want to startle the unwary. At other times, the transition may itself be the crucial expressive element.

Transitional nodes, when used naturally, can help maintain coherence and causality. But, as noted in the literature on hypertext fiction, apprehension of pattern can itself lend coherence (Harpold 1991; Hayles 2000). Some patterns are more natural to sculptural hypertext than others. For instance, although cycles are the central structural motif observed in most successful calligraphic hypertexts (Joyce 1997; Bernstein 1998b), cycles prove rare in Card Shark. To permit recurrence, the same passage must occur on two or more different cards.

Incoherence, oscillation, repetition, and cycles are sometimes seen as inherent to hypertext, especially by those who don’t like the hypertexts they’ve read (Birkerts 1994; Miller 1998). These patterns, however, may inhere more closely to the inclinations of the artist and the propensities of the tools the artist chose, rather than to the nature of the medium itself.

Social Shark

Reading is often considered a solitary activity, but we might also enjoy Card Shark hypertexts with company. Extending Card Shark for collaborative reading creates Social Shark, and with it some interesting opportunities.

Consider two readers, Mr. Green and Ms. Blue, who meet (perhaps over the Net) to read a Social Shark hypertext together. The computer unwraps a fresh deck, shuffles the cards, and deals seven cards apiece to Green and to Blue. The two readers take turns, following the rules of Card Shark; the reading continues until neither reader can continue.

Each card, in addition to its text, its preconditions, and its assertions, is labeled with a green number and a blue number. Whenever a card is played, Mr. Green receives a number of points specified in green, and Blue receives the number of points specified in blue. At the end of the reading, when neither player can continue, the player with the highest score wins.

Extensions to additional participants are easily envisioned by adding additional score numbers. Alternatively, a third player might seek to maximize the combined blue and green score, a fourth player might seek to minimize it, and so on. Indeed, the game need not be competitive: Green and Blue might conspire toward a shared goal, perhaps in a common struggle against some quality inherent to the fictive universe. In that case, the gameworld itself would constitute a third character.

How might writers assign these values to cards? One simple approach identifies each player with a goal: Green is rewarded whenever Love grows between two characters, while Blue is rewarded when complications or misunderstandings separate them. The goal may extend to plot and subplot, where Green is rewarded whenever anyone falls in love. But the goals of Green and Blue need not be orthogonal: Green might be rewarded when Love advances while Blue is rewarded by anarchy and chaos. In this case, Green’s triumph is a romantic comedy – Pretty Woman or The Tempest – while Blue’s triumph is hilarity – Horse Feathers or The Importance of Being Earnest.


Thespis, our second exotic tool, extends the core idea of Card Shark by allowing many agents to participate in a single hypertext. Each agent or actor receives cards that describe possible actions, and each in turn selects an action to take. One actor represents the reader; the reader chooses actions as she wishes. The other agents are computational structures; they choose according to their design.

Each actor has a name and a simple internal state. Each actor also has its unique function Happiness (state), modeled as a linear combination of state values. Actors select available actions that are likely to improve their happiness. Faced with the same options, different character may choose different actions. One character may value money more than another; one may crave excitement while another avoids it. These crude behaviors are not meant to model psychology, but merely to provide the appearance of intentionality and individuality. We are not making (Mateas and Stern 2001); we’re making theater (Laurel 1991).

Each actor moves across the bounded, two-dimensional space that represents the stage. The reader sees and hears things that happen nearby; more distant actions may be unnoticed. This spatial component neatly reifies the hypertextuality of Thespian space; rather than following this link and not that one, we are standing here, not elsewhere. Perhaps we are sitting at the bar and talking with Hugh, Cathy, and Kaj. Across the room, we might see Randy and Stuart arguing with Susana, but if we want to hear them we’ll have to walk over there – and then we’ll miss the action now unfolding before us.

We may constrain actions in Thespis by reference to the environment and to the context.

It’s getting dark. Winter is coming. I tried to remember winter – the last winter before the war. It seems so long ago.

Without constraint, any actor might say this. But we can easily impose constraints, choosing who may say this, or to whom it may be said. We could specify when and where it can be spoken. A variety of partial constraints are provided; for example, an action that is OnlyExcited can only be performed if the actor is unusually agitated, and a Private action can only be performed if the participants don’t know they’re being observed.

Actors and actions are simple. Rather than create complex actors, we create simple automata that say interesting things about important matters. Actors can move (to a landmark or to another actor), they can use props (eating, for example, if they feel like it), and they can talk. The point of this computational mechanism is merely to keep the actor-automaton from breaking the theatrical illusion. We enforce a naïve physics of the stage, decreeing, for example, that actors should never walk through walls. Simple logic can give rise to complex emergent behavior (Resnick 1997), and this aggregate behavior can be convincingly organic.

Perhaps more important, though, is the recruitment of the reader as a dramatic coconspirator. If simple automata are well-written, if they are engaging and convincing, readers will want to attribute agency, intentionality, and emotional depth to them (Reeves and Nass 1996).

A Thespian Example

What would it be like to read a hypertext written for Thespis? Let’s imagine one. We beg the reader’s indulgence for this lengthy exposition of the plot of an unimportant, incomplete prototype. Demonstrating narrative is a vexing problem; the only way to understand a work is to experience it, and even then the illustration may founder on accidents of taste, interest, or understanding. The intent of this section is to establish the example in sufficient detail to permit the reader to construct a similar hypertext in her own laboratory. The alternative – presenting a formalization of Thespis – seems futile. We’ll call it The Trojan Kids, an experimental adaptation of Euripides’ Trojan Women, in modern dress. It’s set in a large, open, metal-roofed shed, a community center for a small village that has fallen recently to the conquering invaders. It’s the night of the big school dance; life goes on. It could be France in 1940, it could be a village in Rwanda or Kosovo or Chechnya.

The reader is ALICE, uncertain, unsure, unimportant.

She is met at the doorway by EMILY, a plain and unpopular student who has done most of the work of arranging the dance, setting out the refreshments, getting permits from the Provisional Government. She greets us warmly:

Emily: Come in, come in. I’m so glad you’re here. Everybody’s here. Come in, let’s all be together, together again. Let’s celebrate the blessings of peace.

Alice: But, Emily, we lost! After ten years, our gates lie in ruins, Greek soldiers patrol the streets, smoke rises from the palace. What blessing is this?

Emily: Defeat is bitter, sure. But now we have peace! At last! With honor! And in our time. Now we can have our party. It’s our tradition, and the Greeks gave us a permit.

Go on in. Try the shrimp – I hear the dip is really spicy!

Emily is a Pollyanna, an accommodator, a collaborationist in embryo, and the stink of a dark future hangs over her irritating cheeriness. Can she be saved?

Inside, there’s a crowd. They’re kids; most of them have simple motivations. Some hope to get drunk. Some hope for a memorable moment of basketball or Nintendo. Some hope to get lucky. Some of these kids were conceived in the secluded dunes out back, at a party very much like this, just five or six years before the war began.

Others have more to say. CASSIE (Cassandra) is dark, sexy, strange. She knows stuff. She’s seen Emily, for example, a few years from now, her head shaved, hounded through the street. Cassie knows that some of her friends here tonight will be in the cheering crowd. She knows that others won’t make it that far. While Cassie isn’t popular, and nobody pays attention to her stories, she’s hard to ignore; she draws boys like a flame and those boys draw girls. She’s rarely alone.

POLLY (Polixena) is the old king’s niece. She shouldn’t be here; she doesn’t know this crowd; she goes to private school. She was away from the palace when the soldiers came. She’s on the run. She’s escaped the patrols so far, but she’s running out of options. Perhaps, if she can blend in, nobody will notice that there’s a member of the royal family still at large. Perhaps she can stay free, perhaps she can live a little longer.

FRANK and BILL are drinking from a hip flask and debating the relevance of class struggle to the war. They’ve been having this debate since 7th grade. Bill has just realized that he is in love with Polly, that her radiant smile makes the bare 60-watt bulbs burn brighter. He thinks he’s never seen her before, that she’s a new kid. In fact, he’s seen her on TV a thousand times, but not in jeans and a t-shirt. Frank has known for years that he’s in love with Bill, and he sees this immediately and knows that it cannot come to good.

We always begin at the entrance, with Emily, but after that our experience depends on our choices. Perhaps Cassie and her coterie are hovering near the refreshments; we might join them. Cassie has plenty to say (and she can say it, because she’s a prophet and prophets aren’t bound by temporal constraints). Or we might first wander over to Polly. Perhaps we stand off a little ways and eavesdrop as Bill tries to chat up Polly while visions of sand dunes dance in his head. Perhaps Emily rushes up with cups of punch, urging Polly to cheer up and have a great time and get out and dance!

Each of the characters has things to say. They’re kids, they’ll talk about insights and philosophies to anyone who will listen. They move in clusters (as kids at a party do), and sometimes individuals or couples will spin off or two groups will coalesce. Topics of conversation are introduced, old topics are exhausted or discreetly abandoned. Questions and conflicts abound – each individual and unique, but each also connected to the others and to us.

Patterns in Thespis

Is this a hypertext? The Trojan Kids has no links, no blue underlined text, no map view. But these are mere externals. Thespis discloses a chunk of text and then offers the reader a set of choices, and the choice selected determines what is seen next. In Thespis, a reader selects among seven potential options at any moment, although some might, in this specific context, be unavailable. This sounds like a hypertext.

Hypertext, Joyce observed, requires rereading (Joyce 1994; Rau 2000). Thespis can be enjoyed on first reading, but its gamelike qualities encourage rereading and experimentation. In The Trojan Kids, Alice isn’t assigned a mission – there are no captive princesses to be saved – but there are lots of things she could do differently, and many consequences can be imagined. Nothing Alice does can operate by brute force. She can’t protect Polly by fighting off the Army of Occupation with her superpowers; she has no spell to redeem Emily nor elixir to cheer Cassandra. But if she had taken Polly for a midnight walk on the beach, perhaps the police would have missed them in the dark? It’s worth a try. Ineptly done, this is mere puzzle play. Done well, it’s the expression of the tension between tragedy, where fate is inexorable, and comedy, where our effort Or, in romance, our inherent virtue, our intrinsic wonderfulness [Mamet 1998]. can perhaps be rewarded with triumph.

Coherence, causality, and closure – those suspect qualities whose (supposed) absence bedevils the reputation of hyperfiction – can be achieved easily in Thespis if we want them. Assertions form a convenient shorthand for episodes, so explicit temporal contingencies are not difficult to maintain. Overlapping dialogue and episodic collision (Kushner 1993), conversely, provide opportunity for cinematic montage (Miles 2001). We can create Thespian tales that use multiple decks, one after another: a set of characters and actions becomes a scene or an act of a larger drama. Closure, too, can be achieved in all the conventional dramatic ways.

What becomes of the hypertext patterns with which we have become so familiar? Some flourish unchanged; an assertion that opens up a new topic for discussion introduces a split, and the retraction then becomes the balancing join. Indeed, because assertions are easy to retract, split/joins may be more common, larger, and more elaborate in Thespian hypertexts. Basements or mirrorworlds can be constructed in Thespian space; one room contains the theme, another room the counterpoint, with access between the two restricted by a bottleneck. Feints are at once more problematic (we have no maps) and less (characters lie). And tangles, obviously, are Thespis ‘s natural pattern.

Other familiar patterns may be less common in Thespian hypertext. For instance, recurrence in Thespis tends to be rare, brief, and deliberate. Writers who like recurrence (a pattern of cycles) can easily provide duplicate actions, but this is neither as effortless nor as obvious as cycles are in calligraphic hypertext systems. The more elaborate cycles that are the staple of many Storyspace hyperfictions are harder to reproduce in Thespis. Douglas’s cycle, in which repetition signals closure, is difficult to implement, and Joyce’s cycle appears infeasible.

Other Directions

Thespis is a sketch, a prototype. A host of design decisions are arbitrary. Why does the player choose from seven alternatives? Should assertions remain on the blackboard forever, or should they fade over time? When the player sees or hears something, the current Thespis prototype adds the description to the end of a long scroll; would it be better to display the text in a large, dynamic collage (Bernstein 2000)? Or in lots of separate windows? Thespis is a chunky hypertext system that generates a smooth (and linear) text; perhaps it should be generating a smooth hypertext, or a chunky one?

It is also interesting to observe that we can add new actions – indeed, entire new characters – to an existing scene. This could give rise to several intriguing possibilities. Not only might we reread a familiar Thespian play, but we might attempt a reading in the presence of a new, supplemental character. We could envision extensible, recombinant fictions, dramas to which readers could add or remove some of the characters. Or we could let several different people control several characters within the same scene, perhaps through the network; notably, Thespis ‘s constraints on actions averts the worst faults of the graffiti problem, the tendency for open, collaborative writing spaces to be defaced by nonsensical, obscene, or territorial pronouncements.

My Friend Hamlet: Immersion, Games, Interaction

Thespis shares some characteristics with Interactive Fiction (IF) – adventure games, MUDs, and MOOs (Murray 1997). The resemblance, however, is superficial. Unlike IF, the unimportance of the reader’s proxy may be essential to Thespian hypertext.

In IF, the reader is the player, the protagonist, the central character. The reader-protagonist’s actions shape the course of events. In contrast, the reader-protagonist in Thespis is a minor character inhabiting the periphery of the action, a witness to events that unfold. The reader’s choices may indeed alter what happens, but the reader is neither the most interesting nor the most active character on stage.

In most IF, the reader commands a hero protagonist, which naturally leads the reader to test the limits of the possible. That’s what heroes do. The drama rapidly devolves into a negotiation between the reader and the world model: the reader asks to do the unexpected, and the system typically responds with incomprehension. Illusions that place the reader on stage necessarily founder when promised freedom of action is contradicted by the limitations of the simulated environment. Although IF asks us to find a creative, imaginative, and successful resolution to the dramatic problem, the imaginative reader is bound to think of things the creator never envisioned, and the reader’s best thinking inevitably generates the dullest response: “I don’t understand.”

Ironic detachment – the witty, improbable remarks of a Leisure Suit Larry or a Boris Urquhart – makes things worse, not better. If the reader-protagonist still needs to test the rules, any struggle against fate in the narrative reduces, thematically and experientially, to the reader-protagonist’s struggle against the system. This reduction to a gamer’s-picaresque impoverishes interactive narrative, implying that the story of the player’s developing virtuosity is the only story worth attending to. Because the Thespian protagonist is patently unimportant, unheroic, and obviously constrained within the frame of the narrative, dramatic necessity need not continually draw the reader’s attention to the limits of the possible (and hence to the shortcomings of the system).

The computational environment can never match our aspirations. Another way to say this is: we are always more creative than our systems (and that is the good news). Allusions to unlimited computing power of the future (the Starship Holodeck) can’t rectify the fundamental problem: readers will always want to do things nobody (and no computer) could anticipate. That, after all, is why people are interesting, and why we enjoy fiction.

Even if we could experience Hamlet on the Holodeck, it wouldn’t work. Tragedy requires that the characters be blind (as we ourselves, at times, are blind). If you let a sane and sensible reader-protagonist into the room, everything is bound to collapse. Take Hamlet: it’s absolutely obvious that he should go back to school, get roaring drunk, get laid, and await his opportunity. He knows this. Horatio knows this. Ophelia knows this. Even Claudius and Gertrude know – why else send for his college pals? Nobody can bring themselves to say the words – that’s the tragedy. But, if you’re the sane and sensible character with Hamlet on the Holodeck, what’s to stop you? Only brute force and error messages (“You can’t do that”) that call attention to the arbitrary boundaries of the world. If you make Hamlet a game, it has to be rigged so that actions taken by a reasonable and sane reader-protagonist – not to mention a wildly inventive one – do not derail the train of events that must ensue if this is to be Hamlet and not, say, Timon of Athens or A Midsummer Night’s Dream.

It’s not just Hamlet. Oedipus needs to get out of town and change his name, to enter the Foreign Legion or the Witness Protection Program. Antigone needs a long talk with her rabbi. Juliet needs to tell her parents exactly what she did last night. She can’t, of course, but what’s to stop you?

This game is rigged and, more importantly, the game constantly calls our attention to the fact that it is rigged. Whenever we struggle against the bonds of fate (and the boundaries of the system), we’re told, “I don’t understand.” The more we struggle – the more conviction and intelligence be bring to the action – the greater the likelihood that the system will find no appropriate response. The artist’s struggle to convey difficult truths is inevitably superceded by the reader-protagonist’s struggle to do difficult or at least interesting things with a necessarily recalcitrant world-model.

Card Shark avoids this contradiction by foregrounding the familiar convention of reading and drama: we may want our favorite characters to prosper, but as spectators, we cannot choose the outcome. Thespis gives us a greater range of action and might offer us a chance to take a role, but that role is not central and our limitations are evident.

In contrast to the more ambitious interactive fiction projects (Mateas and Stern 2001), Thespis makes no attempt to model character, emotions, or cognitive state. A trivial mechanism lets the actors choose among possible actions. What really matters is what is said, and everything that can be said in Thespis is written in advance. Authorial control retains its customary (if ambiguous) place.

Interactive fictions tend to be spatial; the implicit narrative of Adventure and Myst is one of travel and discovery (Jenkins and Fully 1994). Thespis is performed in imaginative space, too, but Thespian spaces tend to differ in scale and design from adventure-game spaces. IF spaces tend to be numerous, small, varied, and thinly populated. Thespis, on the other hand, uses space chiefly as a place in which actors move; Thespian spaces tend to be large, bland, and crowded. There may be interesting settings and props in a Thespian world, but these are static and durable. The actors, on the other hand, are moving and speaking; if we don’t listen to them now, we may never hear what they say.

Towards Evaluation

Card Shark and Thespis present alternative approaches to hypertext. They stand far afield from Storyspace and the web, but share the core values of literary hypertext.

A difficult question, of course, concerns utility: are Shark and Thespis good for anything?

First, as formal experiments, their simplicity is inherently desirable. Neither system requires elaborate infrastructure. Indeed, Shark was readily prototyped in Flash, while Thespis was a matter of a few weeks’ programming. Variants of either system can easily be implemented and explored without extraordinary technological or financial investment.

Thespis was designed for works of the imagination. Can it be used for argumentation, pedagogy, or technical documentation? Clearly sequential presentation is invaluable for mathematical proofs and for some kinds of schoolwork. Information retrieval is clearly ideal for answering specific, well-posed questions. If Thespian hypertext has a place outside the world of imagination, that place most probably lies in exploring multifaceted topics for an expert, engaged audience.

One can readily envision, for example, a lively Thespian discussion about areas of professional and scholarly controversy. Was Captain Cook considered a deity, or merely an unwelcome dinner guest? Should Web site design emphasize familiarity and ease of use, or unique identity and value? Argumentative strategies, personified as characters, could inhabit a discursive space through which the reader might wander, witnessing debate and pursuing the most interesting parts of the discussion, but simultaneously aware that these discussions necessarily impinge on a wider range of questions.

We might also envision Thespian explorations of naturally discursive subjects – the aesthetics of algorithms, the beauty of chemical synthetic pathways, or the experience of life in London in 1680. Here, the weight of the argument lies in the accumulation of detail and in allowing readers to discover the specific details that speak most powerfully to them. If you want to argue that algorithms or sculpture are beautiful, you’d best be prepared with a variety of examples and let the audience tell you what they like. While this approach to the art of persuasion differs markedly from an argument that proceeds in lockstep from proposition to lemma to conclusion, intellectual history is rife with examples of precisely this more capacious rhetorical artistry, from Burton’s Anatomy of Melancholy to Kolb’s Socrates in the Labyrinth.

Wandering through Thespian spaces lets the reader see what she wants, yet Thespis also indicates unobtrusively that there is more to see, and sketches where she might go next.


Ken Perlin responds

Andrew Stern responds

Mark Bernstein and Diane Greco respond


Eric A. Cohen, Mary Cavill, and Charles S. Bennett read early drafts of this paper and made many valuable suggestions. Portions of this paper originally appeared, in somewhat different form, in Proceedings of the Twelfth ACM Conference on Hypertext and Hypermedia, Hugh Davis, Yellowlees Douglas, and David G. Durand, eds.

References: Literature

Aarseth, Espen (1993). “Nonlinearity and Literary Theory.” In Hypertext and Literary Theory, edited by P. Delany and G.P. Landow. Baltimore: Johns Hopkins University Press.

Alexander, C., S. Ishikawa, M. Silverstein, M. Jacobson, I. Fiksdahl-King and S. Angel (1977). A Pattern Language: Towns, Buildings, Construction. New York: Oxford University Press.

Amerika, Mark (1997). Grammatron.

Arnold, Mary-Kim (1993). “Lust.” In Eastgate Quarterly Review of Hypertext 1, no. 2. Watertown, MA: Eastgate Systems.

—., and Matthew Derby (1999). Kokura. Watertown, MA: Eastgate Reading Room.

Bates, J., A.B. Loyall and W.S. Reilly (1992). “Integrating Reactivity, Goals, and Emotion in a Broad Agent.” In Proceedings of the Fourteenth Annual Conference of the Cognitive Science Society, Bloomington, Indiana, July 1992.

Bernstein, Mark (1991). “Deeply Intertwingled Hypertext: The Navigation Problem Reconsidered.” Technical Communication.

—. (1998a). “Hypertext Gardens.”

—. (1998b). “Patterns of Hypertext.” In Proceedings of Hypertext 1998, Pittsburgh, PA.

—. (2000). “More Than Legible: On Links that Readers Don’t Want to Follow.” In Proceedings of Hypertext 2000, San Antonio, TX.

—. (2001). “Card Shark and Thespis: Exotic Tools for Hypertext Narrative.” In Proceedings of the Twelfth ACM Conference on Hypertext and Hypermedia, edited by Hugh Davis, Yellowlees Douglas, and David G. Durand. New York: ACM Press.

Birkerts, Sven (1994). The Gutenberg Elegies. Boston: Faber and Faber.

Bly, Bill (1997). We Descend. Watertown, MA: Eastgate Systems, Inc.

Bolter, Jay David (1991). Writing Space. Lawrence Erlbaum Associates.

—. (1997). “The Rhetoric of Interactive Fiction.” In Texts and Textuality, edited by P. Cohen. New York: Garland.

—., and Michael Joyce (1987). “Hypertext and Creative Writing.” In Proceedings of Hypertext 1987, Chapel Hill, NC.

Chapman, Wes (2001). Turning In. Watertown, MA: Eastgate Systems, Inc.

Coover, Robert (1992). “The End of Books.” The New York Times Book Review, June 21, 1992: 23-25.

—. (1993). “Hyperfiction: Novels for Computer.” The New York Times Book Review, August 29, 1993: 1-12.

Coverly, M.D. (2000). Califia. Watertown, MA: Eastgate Systems, Inc.

Cramer, Kathryn (1993). “In Small & Large Pieces.” Eastgate Quarterly Review of Hypertext 1, no.3.

Douglas, J. Yellowlees (1993). “I Have Said Nothing.” Eastgate Quarterly Review of Hypertext 1, no.2.

—. (2000). The End of Books – or Books Without End? Ann Arbor: University of Michigan Press.

Eisen, Adrienne (2001). What Fits. Watertown, MA: Eastgate Systems, Inc.

Eskilinen, Markku, and Raine Koskimaa (2000). Cybertext Yearbook 2000. Jyvaskylä: University of Jyvaskylä, Research Centre for Contemporary Culture.

Gaggi, Silvio (1997). “From Text To Hypertext: Decentering the Subject.” In Fiction, Film, the Visual Arts, and Electronic Media. Philadelphia: University of Pennsylvania Press.

Golovchinsky, Gene, and Catherine C. Marshall (2000). “Hypertext Interaction Revisited.” In Proceedings of Hypertext 2000, San Antonio, TX.

Guyer, Carolyn (1992). Quibbling. Watertown, MA: Eastgate Systems, Inc.

Harpold, Terry (1991). “Threnody: Psychoanalytic Digressions on the Subject of Hypertexts.” In Hypermedia and Literary Criticism, edited by P. Delany and G.P. Landow. Cambridge, MA: The MIT Press.

Hayles, N. Katherine (2000). “Flickering Connectivities in Shelley Jackson’s Patchwork Girl: The Importance of Media-Specific Analysis.” Postmodern Culture 10, no.2.

Holeton, Richard (2001). Figurski At Findhorn on Acid. Watertown, MA: Eastgate Systems, Inc.

Jackson, Shelley (1996). Patchwork Girl: by Mary/Shelley/and Herself. Watertown, MA: Eastgate Systems, Inc.

Jenkins, Henry, and Mary Fully (1994). “Nintendo and New World Narrative.” In Communications in Cyberspace, edited by Steve Jones. Los Angeles: Sage.

Joyce, Michael (1988). “Siren Shapes: Exploratory and Constructive Hypertexts.” Academic Computing 3 (1988): 10-14, 37-42.

—. (1990). afternoon, a story. Watertown, MA: Eastgate Systems, Inc.

—. (1994). Of Two Minds: Hypertext Pedagogy and Poetics. Ann Arbor: University of Michigan Press.

—. (1997). “Nonce Upon Some Times: Rereading Hypertext Fiction.” Modern Fiction Studies 43, no.3 (1997): 579-597.

Kolb, David (1997). “Scholarly Hypertext: Self-Represented Complexity.” In Proceedings of Hypertext 1997, Southampton, UK.

Kushner, Tony (1993). Angels in America: Millennium Approaches. New York: Theater Communications Group.

Landow, George P. (1990). “Popular Fallacies about Hypertext.” In Designing Hypertext/Hypermedia for Learning, edited by D.J.J.a.H. Mandl. Heidelberg: Springer-Verlag.

—. (1992). Hypertext: The Convergence of Contemporary Critical Theory and Technology. Baltimore: Johns Hopkins Press.

Lanham, Richard A. (1993). The Electronic Word: Democracy, Technology, and the Arts. Chicago: University of Chicago Press.

Laurel, Brenda (1991). Computers as Theatre. Reading, MA: Addison-Wesley.

Malloy, Judy (1993). Its Name was Penelope. Watertown, MA: Eastgate Systems, Inc. First shown at The Richmond Art Center in the exhibition “Revealing Conversations,” (1989) curated by Zlata Baum.

—., and Cathy Marshall (1996). Forward Anywhere. Watertown, MA: Eastgate Systems, Inc.

Mamet, David (1999). True and False: Heresy and Common Sense for the Actor. New York: Vintage.

—. (1998). Three Uses Of The Knife: On the Nature and Purpose of Drama. New York, Columbia University Press.

Mateas, Michael, and Andrew Stern (2000). “Towards Integrating Plot and Character for Interactive Drama.” Working notes of Socially Intelligent Agents: Human in the Loop – 2000 AAAI Fall Symposium Series. Menlo Park, CA: AAAI Press.

—., and Andrew Stern (2001). “Towards Building a Fully-Realized Interactive Drama.” In Proceedings of Digital Arts and Culture, Providence, RI, April 2001.

McDaid, John (1992). Uncle Buddy’s Phantom Funhouse. Watertown, MA: Eastgate Systems, Inc.

Miles, Adrian (2001). “Hypertext Structure as the Event of Connection.” In Proceedings of the 12th ACM Conference on Hypertext and Hypermedia, Århus, Denmark.

Miller, Laura (1998). “” The New York Times Book Review, March 15, 1998: 43.

Moulthrop, Stuart (1991). Victory Garden. Watertown, MA: Eastgate Systems, Inc.

Murray, Janet (1997). Hamlet On The Holodeck: The Future of Narrative in Cyberspace. New York: Free Press.

Nelson, Theodor Holm (1982). Literary Machines.

—. (1987). Computer Lib / Dream Machines. 1st edition: self-published, 1974. 2nd edition: Redmond, WA: Tempus Books/Microsoft Press.

Rau, Anja (2000). “Wreader’s Digest: How To Appreciate Hyperfiction.” Journal of Digital Information 1, no.7 (2000).

Reeves, B., and C. Nass (1996). The Media Equation – How People Treat Computers, Television, and New Media Like Real People and Places. Cambridge, MA: Cambridge University Press.

Resnick, Mitchell (1997). Turtles, Termites, and Traffic Jams: Explorations in Massively Parallel Microworlds. Cambridge, MA: The MIT Press.

Rosenberg, Jim (1993). “Intergrams.” Eastgate Quarterly Review of Hypertext 1, no.1 (1993).

Sondheim, Alan (2001). “Codework.” American Book Review 22, no.6 (2001).

Stern, Andrew (1999). “Virtual Babyz: Believable Agents with Narrative Intelligence.” Working notes of Narrative Intelligence – 1999 AAAI Fall Symposium Series. Menlo Park, CA: AAAI Press.

—. (2001). “Creating Emotional Relationships with Virtual Characters.” In Emotions in Humans and Artifacts, edited by R. Trappl. Cambridge, MA: The MIT Press.

Strickland, Stephanie (1998). True North. Watertown, MA: Eastgate Systems, Inc.

Tolkien, J.R.R. (1984). “On Fairy Stories.” In The Monsters and the Critics, and Other Essays, edited by Christopher Tolkein. Boston: Houghton Mifflin.

Tosca, Susana Pajares (2000). “A Pragmatics of Links.” In Proceedings of Hypertext 2000, San Antonio, TX.

Walker, Jill (1999). “Piecing Together and Tearing Apart: Finding the Story in afternoon.” In Proceedings of Hypertext 1999, Darmstadt, Germany.

Yankelovich, N., N. Meyrowitz, and A. van Dam (1985). “Reading and Writing the Electronic Book.” IEEE Computer 18, no. 10 (October 1985).

Zellweger, Polle T. (2001). “Fluid Annotations in an Open World.” In Proceedings of the 12th ACM Conference on Hypertext and Hypermedia, Århus, Denmark.

Reference: Games

Adventure. Will Crowther and Don Woods (1972-1976)