In his epic Dark Souls Diaries series, Matt “Steerpike” Sakey wrote about a key moment when he felt guilty for killing an NPC he had intended to save. Sakey didn’t have long to mourn. Rather than leave him to wallow in his misery, one commenter told him there was actually nothing he could do. Don’t feel bad about it.
Player guilt is so easily destroyed, it seems, if we learn everything is a foregone conclusion. We are fascinated by what lies behind the curtain and the fear that the game might be making a fool of us, exploiting us through an illusion of agency. No one wants to be Stanley of The Stanley Parable (Galactic Café, 2013), the developer’s puppet.
We crave the weight of consequence yet revel in its destruction. How do we make sense of this contradiction?
the hypertext story
A few months back, I revealed my lack of twine love in an article titled You Complete Me. My particular problem was the conditioning that comes from using hyperlinks day in and day out. I could not stop myself from scurrying through the twine links to find an ending, paying little regard to the words on the way. It was so easy to run amok through a twine that I felt that decisions had no weight. This was never going to be my story.
The implied question: if a player can easily select a particular ending with a few well-placed clicks, whither agency?
In the comments, George Buckenham referred me to John Brindle’s interpretation of twines as “node maps”. Brindle liked to sketch a twine’s cartography, the decision tree through which the player advanced the plot. I had always approached things from the perspective of player agency, what we can and cannot do, instead of looking at the story as a whole.
Mapping the decision tree is like reverse engineering and it’s a useful exercise for a game critic to unravel the blueprint of the game, or a game designer to learn how things are made. Here’s the opening passages of the twine Dive (Amanda Lange, 2013), from which I learnt never to go scuba diving:
It’s common to think of the act of “reading” to be a linear, sequential experience. With this lens, we could visualise a twine as a reader choosing which story tracks to lay before the narrative train. I’m going to take “narrative” to mean a particular story path through to the end.
Suppose I crack open a twine and the protagonist can either drive to an important destination in a brand new red Ferrari or travel by foot. I choose the car – but it’s booby-trapped and explodes.
I could see this as a fail state, but there are plenty of twines which do not have anything resembling a “success” state. I know we’re not done because the story feels truncated, incomplete. I don’t need to be told this.
Twine allows us to rewind such choices, but some twine developers do away with this ability – Duck Ted Bundy (Coleoptra-Kinbold, 2014) and The Matter of the Great Red Dragon (Jonas Kyratzes, 2014) are two examples that do this from the recent Fear of Twine exhibition. However, even if the reader is unable to rewind, it is always possible to restart.
There’s obviously more to discover in this twine, so I rewind and choose the walk option instead. Eventually I step through the game to an ending passage which feels more complete. Yet it turns out the walk option takes us through a narrative that never refers to a car bomb, so a reader who selected the walk option would be oblivious to its existence. But I cannot “unsee” those future paths already traversed, so I am aware of a car bomb even though it is never referenced during the walking branch of the story.
There’s a significant difference between the narrative I selected and the richer model of the story I’ve assembled in my mind. I do not think of the car bomb death as the narrative of the game neither do I think of the successful ending as the narrative. I remember both: a story of a car bomb which is possible to survive and reach success. I have built a node map:
By exploring every branch, I acquire a composite understanding of the story that’s independent of a single narrative. We are so comfortable with the concept of multiple narratives that we rarely distinguish between “an instance of narrative” and the complete picture, the hypertext story.
Twines are sometimes snubbed as digital “Choose Your Own Adventure” (CYOA) books yet there some important observations in this comparison.
In truth, CYOA books were often in thrall to the idea of finding the correct narrative rather than choosing your own. The stronger the emphasis on correct over personal, the more irritating the book. In the classic Fighting Fantasy book Deathtrap Dungeon (Ian Livingstone, 1984) the hero must find several gems scattered about a dungeon. The book is much beloved but if the reader reached the end of the dungeon without the necessary gems, the story came to an abrupt end. Forced replays like this have several repercussions.
One, it means CYOA books, hypertext works of the 80s, could only be finished if a reader developed a significant repository of story information from previous attempts. In other words, they were building a composite picture of the hypertext story, not “choosing their own adventure” at all.
Second, being tripped up time and again by sudden, unpredictable deaths meant some readers shifted from making brave choices to checking out all the options – using their fingers as rewindable bookmarks, allowing them to back up a few passages in the text. Sometimes, players might re-roll the dice in combat a few times to “encourage” a positive result if they were frustrated.
Ultimately, I ended up sending dozens of adventurers to horrible fates just to get a rough idea of what I'd have to do to emerge victorious. I wasn't trying to win, even when I rolled Skill 12 characters; I was just mapping. I knew I stood no chance of victory without loads of foreknowledge. This reduced the book from an adventure to a mechanical exercise, which is a shame. It was also rather displeasing that even after all this work, there was no easy path available -- it's very frustrating when you have mapped every inch of a gamebook and still can't win without a certain amount of dice-fudging.
So the experience of twine and the CYOA book are, in fact, quite similar. The readers of CYOA books were true hypertext explorers, just like the readers of twine. But there’s often a tension between “my personal adventure” and completing the story, finding the One True Ending.
This obsession with a real ending has been ignored by many interactive fiction works. In parser-based interactive fiction Photopia (Adam Cadre, 1998) – which is actually closer to twine than parser – you can replay to have different conversations with central character Alley and it is through these conversations you can learn more about her. There is only ever one ending to the story, but a single play will not tell you everything.
We’re used to the common wisdom of books and films being uni-directional media. We start on page one and know we’re finished when the credits roll. We’re drawn to the idea of “The Narrative”, a master sequence of events being played before us. Even films like Inland Empire (David Lynch, 2006) or 21 Grams (Alejandro González Iñárritu, 2003) which present fragmented stories, still take their passengers on a directed ride from start to finish. We’ve bought into the idea that storytelling is about a journey to an endpoint, an authored destination.
The real destination is the creation of meaning, whether that be the reader’s interpretation or reconstructing the author’s intent. The work is not completed by reading the final page but by reading the all of the pages. We should be mindful of attributing holy status to the raw text as our interpretation is influenced by sources outside of it, such as what we know of the author, other related works and conversations about the work. (The academic terms you may be grasping for are intertexuality and paratext.)
Hypertext tales are often sold as games of choices, of agency, of your version of the narrative – when it would be better to embrace what they are good at, which is telling a different kind of fiction: one which is host to multiple narratives, narrative superposition for want of a better term, from which meaning can be derived.
And it’s probably about time I stopped pretending I’m writing about twine fiction.
everything is hypertext
Virtually all narrative-driven videogames offer hypertext stories. Players work their way from one story node to another not through clicking, but through sections of gameplay. Sometimes the node branching is more explicit through the use of dialogue trees. What has been bothering me for a while is that despite all the cheerleading about choice and personalisation in our game stories, we’ve been embracing them as works of hypertext for years.
Let’s start with the game critic. Critics are expected to be aware of the significance of the choices in any game, to show some effort to understand how the story works. Although I haven’t read Brendan Keogh’s book on Spec Ops: The Line (Yager Development, 2012) Killing is Harmless nor Rowan Kaiser’s as-yet incomplete Mass Effect (Bioware, 2007) book Possibility Space, I imagine both works explore the decision trees of those games and what they imply; these critiques are hypertext-literate.
But it’s not just critics who are exploring the hypertext nature of narrative-driven games. Going back to the ending of Deus Ex (Ion Storm, 2000), the game offers the player three choices. Each one leads to a different challenge with a different closing cutscene. The primary story branches as follows:
The Ion Storm team wanted to offer three starkly different futures, allowing the player to reflect upon what the game had taught them. The question is which one did players choose? I can tell you what players chose. They chose all bloody three.
Obviously players, who have come this far, have a desire to exhaust the game’s content; it’s just too tantalizing to leave the other two endings fallow after choosing one. Seeing one ending wasn’t considered “complete” – we had to see all three, and that’s hypertext fiction in action.
It’s important to realise that it’s not all about exhausting content because if that were true we wouldn’t spend so much time following up the alternative endings on YouTube and walkthroughs. A couple of years ago I described my dissatisfaction with Lone Survivor (Superflat Games, 2012) in Destroy After Playing. In Lone Survivor, it is unlikely a single playthrough will lead to narrative closure and a player will need to play through the entire game again to glean more information about the game’s plot. What did I do? I went straight to YouTube to do the job for me.
The Stanley Parable is about the relationship between player and developer but most coverage interpreted the game as being about choice. Parable lampoons choice and is the definitive hypertext game, demanding players track down every ending for laughs. No one plays Parable for the “good ending” or a personalised playthrough, they seek its meaning in aggregate.
In fact, the different narratives of The Stanley Parable are contradictory. There are many other works which also embrace this such as Fear of Twine’s The Girl in the Haunted House (Amanda Lange, 2014) and interactive fiction Aisle (Sam Barlow, 1999) in which one move not only determines what happens but also what happened. We appreciate them as a whole and not a single story. We do not reject these experiences for being inconsistent because it is part of their explicit design.
the retread problem
However, Aisle is the only example here that sidesteps the fundamental flaw in hypertext fiction. Katz hints at it in his Deathtrap Dungeon review, but Espen Aarseth executes a more rigorous take-down in his book Cybertext: Perspectives on Ergodic Literature (Aarseth, 1997).
After the player has navigated through a hypertext story once, it is inevitable that the player will be forced to revisit previously trodden sections in an effort to find virgin ground in the text. It’s a wearying exercise and a subtle change in a familiar looking passage is not going to be enough to convince a player’s bored eye to do anything other than jump on a link.
This is why developers try to shoehorn every important story beat into a single playthrough, exposition and cutscenes be damned. They realise it’s not necessarily fun to jump through the same hoops time after time trying to find a little bit more of the story.
Where are the tools for reading hypertext fiction?
Most games provide scant support for hypertext readings. It’s become popular with some indie games to signal how many endings have been reached; the acclaimed Papers Please (Lucas Pope, 2013) tucks away a series of unremarkable dots in the corner of the start game screen to denote which endings the player has visited.
The purpose of this mechanism is to inform players of “undiscovered content”, to make them feel that a game is far from complete, despite reaching a completion state. Yet this is also a kind of hypertext navigation device: dear player, you have not yet read the whole story.
There’s also the “chapter selection” menu, which allows a player to jump directly to a particular point in the story, but this is usually reserved for games that offer a single narrative thread with no branching such as Half-Life 2 (Valve Software, 2004) and Bioshock Infinite (Irrational Games, 2013).
The problem is that hypertext literacy opposes the conceit of narrative-based agency. The easier it is to navigate the decision tree, the more obvious it becomes that players aren’t making “real decisions”. Developers employ different tricks to distance the work from a hypertext perspective, to make decisions seem poignant and weighty.
The most effective way to achieve this is to create a hypertext structure which is so dense that it dulls our need to explore it – a convincing discrete approximation of free will. But this is expensive and complex work, particularly in AAA development where branches have a heavy cost in artwork and motion capture. Indeed, developers increasingly want to obscure the story scaffolding which leads to mixed results such as Beyond: Two Souls (Quantic Dream, 2013) being tarred a “linear” game when in fact it features plenty of branching. Brian Boudreaux wrote: “Players had no clue that they were making choices, or that they missed whole scenes until they talked to another player, or replayed the game.”
There are cheaper alternatives: demoting save games to checkpoints, stretching the distance between cause and effect to make replaying consequences more prohibitive and, down at the extreme end, eliminating all save game mechanisms or even revoking the ability to replay. For all the talk of games with “real choices,” developers have typically dealt with this need by stripping the player of narrative control.
Let’s put aside developers’ lacklustre support for hypertext navigation, because the solution to the hypertext retread problem has always been here. It is the collaborative reading.
Mapping hypertext is often laborious work for the individual, but not for the group. We’ve always taken part in a grand collaborative read but social media, YouTube and organised mapping projects like wikis have accelerated hypertext understanding. These elements of paratext have become part of the reading process.
Despite the obfuscation efforts of developers, players have been painstakingly mapping these decision trees. Look at the walkthroughs. Look at the wikis. Look at the YouTube videos.
The hypertext maps are out there.
It is important to separate hypertext storytelling from other forms of storytelling.
Consider environmental storytelling which I last discussed in The Beautiful Dead which is heavily authored but normally renders a static, dead backstory – not one which the player has any say in. For example, the environment of Bioshock (Irrational Games, 2007) expresses what kind of city Rapture was and also intimates that something terrible happened to it.
There’s so-called emergent storytelling, which are take-home stories that describe what happened in play. Take Matthew Weise’s wonderful journal of his time in ZombiU (Ubisoft Montpellier, 2012) or my encounter with The Beast of Miasmata (IonFX, 2012). These were experiences that arose from the structure of play. I do not expect emergent storytelling to deliver the layered hypertext writing of Planescape: Torment (Black Isle Studios, 1999), but neither do I expect hypertext fiction to surprise on repeat viewings like Spelunky (Mossmouth, 2012). Arguing whether one is better than another is a moot point.
The Holy Grail, of course, is procedurally-generated story, emergent stories that look just like authored ones. Recently, Ken Levine gave a GDC presentation on a “narrative lego” approach but I remain unconvinced. Anyone can describe a concept of replayable story; the difficulty is in implementing something that convinces the way authored story does. Okay, Symon (GAMBIT, 2010) and Stranded in Singapore (GAMBIT, 2011) have procedurally-generated stories; but all that “narrative lego” feels predictable after just a couple of throws of the dice. Does this have any more longevity than authored hypertext?
This shines more light on the issue of game ethics I had with Dishonored (Arkane Studios, 2012): the truth is no one cares about the ethical conundrum about whether to take lives or use stealth. The use of violence was more about an alternate form of play and different game endings. It’s as much a hypertext exploration as it is playing a different kind of game. The idea of feeling the hideous weight of those consequences is way, way down the list of wants, here.
This isn’t to say there aren't plenty of examples where the hypertext branches transmute into something that feels like agency. Many people found the choices they made in Papers, Please uncomfortable and the mistakes players made in Cart Life (Richard Hofmeier, 2012) affected some deeply. Those who enjoyed the Mass Effect trilogy and The Walking Dead (Telltale Games, 2012) cannot help but talk about the consequences of their choices. It is clearly possible to synthesise a sense of complicity in these stories.
It is as much a player’s responsibility as it is a developer’s to maintain immersion. But when choices forced on the player are designed to make the player feel bad, sometimes players call this out as cheap exploitation. But does it always have to be about the player? It might just be the only narrative step consistent with the hypertext story.
And even if we do buy in, it can so easily come undone as I described in Sakey’s Dark Souls encounter. The best we can do is avoid the hypertext reading – I, for one, refused to replay Cart Life multiple times, for fear of undermining how the game made me feel.
But, on the whole, agency in story is a remarkably fragile thing. It also makes little sense in a world where we can excavate alternative narratives and endings with a quick Google search. What players really seek is a work which is complete in a hypertext sense, that the story is closed.
It might be time to forget about forcing players to cry over consequences and let them cry over story.
the hypertext holodeck
Except… well, good luck on that, because players are going to hate you. Implicitly, we’ve been accepting of hypertext but the holodeck framing (players are the real author and should be able to control the story) has a stranglehold on the development of videogame storytelling. Before we move on, I fully acknowledge I write this as someone who has often celebrated the power of consequences.
Players have always pined for additional narrative options, to reflect what they would have done. Even in a game as flat and linear as The Last of Us (Naughty Dog, 2013) there have been calls for alternative choices at the end of the game. It is always difficult to patch in extra choices to serve player whims because they need to be consistent with the existing story structure. It is to satisfy player curiosity at the expense of believability and hypertext closure. It’s the same problem that Tom Bissell complains about in his piece on L.A. Noire (Team Bondi, 2011):
As everyone who plays video games knows, masking systems can be greatly amusing to test and prod, and the first thing I did in L.A. Noire was drive my car directly into some pedestrians and plow through a few streetlights, after which I insisted on driving my partner and me to our first crime scene in a dump truck. Once we got to the crime scene, I stranded my partner there and took off, still in my dump truck, spreading more mayhem. Had this been GTA or Red Dead Redemption, the law would have come calling, but the most incisive criticism I got from my partner during all of this was: “What are you doing?”
But the games still sell despite critics hurling around mouthfuls like ludonarrative dissonance. Do players throw their controllers in disgust if the shooting is fun but the story is cranky or broken? What the player can do has always been more important than what the game wants to say. Players want the asshole option. Even if the asshole option isn’t provided there’s always someone who will make it, capture it and post it on YouTube.
Also, there’s a complex interplay between our holodeck cravings and hypertext structure. What if you embrace the holodeck and throw away the hypertext signposts? You get the Beyond: Two Souls problem of “hidden choices.” Brian Bourdeaux takes this point further:
Players were just not ready for the kind of responsibility Beyond puts on them to better their own experience. When a game does everything for us, we complain that it doesn't give us the freedom to express our agency in the story. When a game gives us the freedom to express our agency in the story, we complain because we don't know how to use it.
There’s another issue, though, with an emphasis on hypertext. Those “sections of gameplay” that bridge the narrative nodes define how those nodes are perceived.
Failing to be successful in Cart Life makes the endings more painful to bear. Even the emptiness of grinding in RPGs becomes part of the journey up the narrative mountain. Instead of clicking through a few twine nodes to the end of the story, the player suffered for hours killing mice and sheep – and this becomes part of the sense that the player suffered with the protagonist for the Great Quest.
That is partly why there was anger over Jennifer Hepler’s suggestion of skipping combat in a game to progress the story. In acknowledging these games were fundamentally hypertext she also implied the grind up the narrative mountain was not important. This casts action sequences as acts of pure masochism instead of one that informs the journey. (The resulting personal attacks on Hepler, of course, were unforgivable.)
Raph Koster wrote how narrative's only purpose in a game was player feedback. Tom Jubert, the writer on Penumbra (Frictional Games, 2007) and The Swapper (Facepalm Games, 2013), was once told that “plot is gameplay’s bitch”. Except it’s possible to turn this on its head. Gameplay is often what gives the narrative some resonance, that gameplay is plot's bitch. Everyone is actually playing a version of Cart Life, except those routine mechanics – point, shoot, point, shoot – are skinned well enough to feel fun.
So where does this leave us?
Viewing games rich with authored story as hypertext fiction is useful for explaining some of the apparent contradictions in player behaviour. There’s obviously more that can be done on the development side to support player exploration of the embedded hypertext. However, there are problems associated with exposing it for easy exploration, as well as burying it completely.
Will there be any change in how we perceive such games? Will we see them more as hypertext in future?
I didn’t think much of Tale of Tales’ Over Games presentation in 2010, but now looking back I see much in common with how I feel about videogame design today. Plenty of people are okay with the lack of choice in Dear Esther (The Chinese Room, 2012) and Thirty Flights of Loving (Blendo Games, 2012): seeing something more fundamental through the “agency illusion”. And some find the well-received albeit flagged choices of The Walking Dead quite hollow.
Things change. Things do change.
notes: alternative endings
- Wired recently ran an article titled Why No One Clicked on the Great Hypertext Story. “It turned out that nonlinear reading spaces had a problem: They were incredibly difficult to write.” Well, that’s not true, Wired. Everyone was looking in the wrong place for hypertext works, imagining them to turn up in bookstores. They should’ve been checking out videogame stores; also consider Twine as a more recent resurgence of the naked hypertext format. I think the real problem is that a hypertext story is difficult to read (the retread problem).
- I used Espen Aarseth’s Cybertext to sharpen my thoughts. It’s a strangely inaccessible book if you’re not familiar with academic nomenclature like semiotics. However, it is interesting and makes a stab at generalizing the notion of a text out to a superset called ergodic literature. Hypertext is just one entry (which Aarseth savages) inside that superset.
- Cybertext is notable as it explains Aarseth’s stance against literary theory and narratology, with concerns that the field will be “colonised” with theories requiring patches and extensions rather than a better more general theory of text (ergodic literature).
- Cybertext would label the hypertexts discussed in this essay “exploratory hypertexts”. The other type is the “constructive hypertext” which permits users to modify the hypertext.
- The melding of hypertext and gameplay into a single “entity of fiction” that I left without a label is fully generalized by Aarseth’s idea of a “textual machine”. The textual machine consists of three elements: operator, verbal sign and medium. That's as much as you're getting out of me on the subject.
- This essay is not an end, it is a means to provoke. Please feel free to throw in counterexamples or highlight flaws. I cover a lot of ground here so I'm going to be stretched thinly somewhere, most likely over France.
- Rare is the game that prevents replays, but they do exist. For example, One Chance (Dean Moynihan, 2010) which I discussed in The Consequences of Consequences.
- Twine works do not necessarily have to conclude. They could keep going in a story loop and the loop itself could hold meaning.
- Twine works can use variables which, if used carefully, can substantially complicate node analysis. State variables will not prevent the game being mapped into a tree structure, but each variable can increase the number of nodes required to explain the game. Usually the node map would be augmented to take account of states rather than simply bifurcating the tree with each variable change.
- It’s possible to argue that “narrative” should cover the full journey the player takes, going forward and backward through all the branches in a hypertext story, thus discarding the possibility of superposed narratives. I’m not sure that’s altogether useful and smudges some boundaries which are better left intact.
- From my hypertext fiction perspective, I would prefer to call a twine a story rather than a “game”. The word game carries this cancerous seed of challenge, of defeating the work. It’s why I asked developers the question of whether they would prefer an alternative to "game" in No Alternative because I am precisely the kind of person who feels uncomfortable writing a work of hypertext fiction and calling it a game. Note there are more complex twines that endeavour to support more traditionally game-like elements.
- Environmental storytelling can support dynamic elements and traces of player activity (e.g. bullet holes) but it’s better suited for world-building and backstory.
- On narrative lego / replayable narrative: is replayability a worthy goal? This is something the gameplay-first crowd always get carried away with. Steady on, fellas. I’ve only read Greg Bear’s Eon once and I thought that was an awesome book.
- Symon and Stranded in Singapore were part of a series of experiments in procedurally-generated story run by Clara Fernández-Vara. I’m only using them as illustrations. The project that followed these, The Last Symphony (Shoefish Games, 2012), was far more interesting and geared towards a different approach to story (indexical storytelling).
- John Brindle also wrote a piece called “Fuck the Holodeck”. It’s got nothing to do with modes of storytelling but you might like it anyway. It’s a rejection of the conventional wisdom that videogames’ goal should be immersion/realism.
- The thoughts in this essay were being developed as I was writing my twine, Truth is Ghost (Joel Goodwin, 2014). I’ll relate how this impacted my twine during the post-mortem next week.
- Actually, that NPC could have been saved.