Writing Bastion

On September 2, I officially joined the small team at Supergiant Games as their creative director, and together we showed our game Bastion for the first time at PAX in Seattle. The response was almost overwhelming, and on a personal level it was one of the most rewarding moments of my professional career. Part of the reason for this is that Bastion is a pure expression of many ideas that are close to my heart – ideas about games, stories, and other things that matter to me – so the enthusiastic response really meant a lot. In the game you'll find a lot of the stuff I've been writing about on this blog put into practice. Granted, my contributions to the project are only just beginning in earnest, but because I was involved in developing the original concept back when my colleagues and I parted ways with Electronic Arts in August of last year, I feel much closer to this game than any other thing I've worked on. Moving forward, I'd like to use this space from time to time to talk more about my thought process while developing Bastion, and to begin with I wanted to explain how we're approaching the game's use of storytelling through narration. But first...

GDC Online Lecture: Delivering Exposition in Games
All this is tangentially related to my GDC Online presentation I gave on Tuesday, October 5. I've posted the slides here and invite you to take a look: I Don't Want to Know: Delivering Exposition in Games. The slides are fully annotated so just by flipping through the slides you'll get a feel for the substance of the talk. If you have a chance to look it over, let me know what you think.

Narration in Bastion
In film, narration is one of the most misused and mood-killing techniques out there, for its unique ability to eliminate the type of ambiguity that adds richness to scenes and characters. While I've often fantasized about being able to read people's minds as a superpower, if movies have taught me anything it's that knowing people's inner monologue would make life far less interesting for someone as neurotic as me.

Nevertheless, Bastion uses real-time narration extensively. Its purpose is to deliver story and exposition, and to build atmosphere, investment, and immersion in close partnership with the gameplay. The narration wasn't part of the original game concept. It was born in a flash of inspiration (through a development process that enables such happy accidents to occur), stemming from a couple of self-imposed constraints. The first constraint was to never interrupt the play experience for the sake of story or for any reason, which meant no cutscenes, no dialogue trees, no pressing the A button to advance through dialogue, and none of the other such trappings that tended to slow the pace of other RPGs. I love many games that do these things, but Bastion just isn't this kind of game. One of the things I miss about games in general is that sense of immediacy that console games used to have (before disc-based media ushered in a new era of loading times and cutscenes), where you'd hit Start and, indeed, start the game. Bastion is meant to be that kind of game. Text-based dialogue wasn't going to work. The team's suspicions about how it would negatively affect the pace of the game turned out to be accurate.

From the outset, before the company was even formed, I wanted to work on a game with some narrative substance and emotional depth, to create an original world with its own characters. We would have these long late-night conversations about how to deliver story in ways only possible through the medium of gaming, because why not? Games should aspire to be games. Cinematics interrupt the play experience no matter how well crafted they are. And as much as I love stepping through dialogue in games like Fire Emblem or Torment, I had to agree that reading lots of text in a game usually isn't a good feeling. All the theorizing needed to be grounded in reality due to the would-be team's small size and limited bandwidth in art and animation. This other constraint meant no elaborate scripted scenes or silent emotive storytelling as in games like Ico or Limbo, where nuanced animation is essential to mood-setting and atmosphere.

Eventually through prototyping and experimentation all this led to the idea of real-time narration, having a narrator who responds to the player's input. From the outset I was interested in having the story begin with a young man rising as if from sleep or from death, to discover a world changed around him in some profound way. The story would start on a mysterious but emotionally low point and expand from there. The intent was to provoke questions for the player immediately, and allow the game to reveal two worlds in parallel: the way things are, and the way things used to be. At any rate, in that waking-up moment, it turns out that just by adding the spoken line "He rose" to coincide with the player's input, it got a lot stronger. (This later changed to the current "He gets up" after further exploration of the narrative style.) This was one of an initial set of lines that our studio co-founder Amir recorded with our audio director Darren and their childhood friend Logan, a theatrical actor who provides the narrator's voice, after Amir suspected that adding narration may bring something positive to the experience. I remember when I first heard it, not knowing what quality it would have, never even having heard Logan's voice before. It felt powerful even inside a low-fidelity prototype. Players don't normally expect this type of output from a game, so it immediately speaks to some of the qualities that are specific to Bastion. The narrator's voice alone says a lot about the game.

The other reason, probably the main reason, Bastion is using narration is because of Logan. In addition to being perfect for the part, Logan offers us one other great advantage: We have access to him. Some people mistook his voice for Ron Perlman's. Let's say we could afford Ron Perlman, lost our minds, and decided he'd be better than Logan for the part. We'd have maybe two or three recording sessions with him for the lifetime of the game. With Logan we can iterate rapidly, and we need to in order to get the narration in the game to feel as closely connected as possible to the moment-to-moment play.

Logan's natural speaking voice is quite different from that of the narrator, though we were always interested in a fantasy-frontier aesthetic, something with some the beautiful-melancholy tone of some of Cormac McCarthy's novels. I also take inspiration from the late William Gaddis, whose novels have characters with such distinctive voices. And so we developed a character who embodies the tone we were interested in. And Logan nailed it.

Bastion's narrator is designed to support our game on a fundamental level. He's a man of few words not only for fictional reasons but also, conveniently, to support a design constraint that we simply can't have him talking a lot during gameplay. Bastion has a very fast feel to it, closer to an action game than a typical action role-playing game. Our narrator needs to be very concise to keep up with the pace.

Five Rules for Writing Bastion
Logan can probably make the stupidest combination of words sound awesome. Even still I'm attempting to write good material for him, in the spirit of not wasting the player's time with bloated unnecessary prose. By exploring the character and which types of narration work best for the game, we gain a low-level understanding of the narration in addition to our high-level goals for it. As such, here are the factors I bear in mind when writing for the game:

1. Dialogue is for subtext. The player's actions in the game are the "text", the surface-level things that happen. When the player explores, builds things, attacks things, or acquires items, these are all clear and affordant actions. There was initially a temptation and a novelty in having the narrator declare these actions along the way. But this would mean missing the point of why we have the narrator in the game. This would have resulted in that brand of movie-style narration I dislike so much.

Our narrator deepens the player's interactions by saying something about them that the player could not have known. He provides character intent, subplot, and backstory through his comments. The ultimate goal with every line is for it to tell you something about the player character, the narrator, the way the world is, and the way the world used to be. For example, the first item you find in the game is a large sledgehammer, to which the narrator says, "Kid finds his lifelong friend." You can see that it's a hammer so you don't need the narrator to point that out, but through the narration you can deduce that the protagonist has history with this hammer and that the narrator knows it. Further, through the narrator's particular delivery you sense that this history has had its ups and downs. Using this type of narration, we gradually build the backstory in the context of the player's immediate actions and surroundings – I would never expect you to care about something that wasn't onscreen. Following the points in my GDC presentation, I mean to deliver on the major questions raised in the game, but moment-to-moment this type of narration should give a sense of a detailed world that existed before you started playing the game.

2. Keep it short. Our narrator is a storyteller but a terse man. Unlike me he doesn't waste his breath, and that's fortunate because our environments are packed with detail and leave no room for long speeches. In order to achieve the moment-to-moment reactive feel we want from the narration, the lines have to be short. Our narrator has a flair for the dramatic and speaks in a low flat voice, so tonally the lines tend to fit well together even if there's a lot of silence in between statements. These were factors in the character design.

3. No breaking the fourth wall. One of the most exciting aspects of having a narrator in our game are all the opportunities to break from player expectations, and raise a lot of interesting questions over time. A temptation in all this is to have the narrator address the player directly or step out of the story and break into metafiction, maybe tutorialize the game by telling you when to press and hold the X button and so forth. But it didn't take long to realize this wasn't going to work. As mentioned earlier on, our high-level goals include building immersion and investment. If the narrator were to break the fourth wall, we might get a momentary gag out of it but we'd be undermining the experience we want to achieve by violating the player's expectations around the game's own logic. We have a cleaner way of training players, and as with everything else, the narrator is there to reinforce those moments without stepping on them.

4. Reward experimentation and playing with finesse. Our narrator provides a great reward system, provided players like what he does for the game. I want players to develop a relationship with him as a character and to feel like they can provoke certain types of comments out of him. This happens to be well in line with the type of play experience we want to deliver, one where players feel like they can do whatever they want in the world, experiment with all the different systems and weapons, explore a bit off the beaten path, and so on. Having the narrator specifically acknowledge these moments tends to provide positive reinforcement in a natural way. We realized the closest thing to what we were going for were some of our favorite announcers in games from completely different genres, from the announcer screaming "BOOM-SHAKALAKA" after an awesome dunk in NBA JAM to Shao Khan saying "Excellent" after a ferocious uppercut in Mortal Kombat (both games were done by Midway in their glory days). The key difference is our narrator isn't quipping, he's telling a contiguous story for the most part. Having him sneak in a few incidental remarks based on the player's choices or performance helps make the whole thing feel personal and specific.

5. No repeats. When done properly, our real-time narration starts to take on the quality of a story unfolding, and starts to get at those high-level goals we want to achieve. But nothing sucks the momentum out of the game's narrative like a repeated line. Almost every game uses repetitious dialogue even if it's got tens of thousands of lines of dialogue in it; combat encounters will repurpose enemy battle chatter and so on. With Bastion we realized that the moment any line repeated itself – for example, our narrator has different things to say if the player falls off a ledge – immersion is broken. You realize in that moment that you're playing a game where the narrator might loop through a host of different lines after a specific event, as in a real-time strategy game where your units will cycle through several responses whenever you issue an order. So we drew a line in the sand: No repeats in the game, not unless you replay the game from the start or restart a scenario from scratch (and even then we mix up the narration). This posed certain design challenges, such as what happens if a player revisits certain areas, but we're happily taking those on in the spirit of maintaining the feel we're going for.

Conclusion
If there's one main underlying point in all this, it's that everything we're doing with the narration in Bastion is there only to support the specific type of play experience we're making. Everything from how the narrator character sounds and how he talks came about purposefully as part of the exploration around gameplay concepts and game themes. Bastion is hardly the first game to use narration to deliver story, so we never set out to pursue the idea of having a narrator purely for the sake of being different. Instead, we're pursuing it because we realized it worked well for the game we wanted to make and for the process we're using to make it.

Narrative Recognition

The narrative of a game should exist only to make the act of playing feel more meaningful, by giving context to the game's systems and scenarios. Even a simple narrative implementation can make a game feel more significant. The simplest and best example I can think of is Space Invaders, the iconic 1978 arcade game whose entire narrative is so conservative, it's limited to the two words of its title. The high concept of defending Earth from evil aliens combined with those expressive-yet-abstract shapes encroaching toward the bottom of the screen to create what's probably the world's first "epic" videogame. But if you took Space Invaders and changed only the name, maybe to something more literal like "Shoot the Sprites", the high stakes wouldn't have been there and players' imaginations wouldn't have run wild from it. The mechanics would have still been great for their time, but I think the narrative is what brought them to life. Space Invaders isn't a typical example of a game narrative, though there's at least some fictional context there, whereas a lot of games have none. It's OK for a game to have little to no narrative, it just better have some truly outstanding mechanics. Tetris might not have needed narrative, but Myst and Puzzle Quest did.

For these and other games that choose to provide the player with a fictional context, one of the best techniques they can use to draw the player into the experience in a profound and memorable way is to anticipate the player's emotions and expectations about the play experience itself, and recognize them at appropriate times through narrative feedback. 'Narrative recognition' is the shortest term I could come up with to describe this simple and practical technique. It does require some daring on a writer's and designer's part, though like everything else we do to avoid mediocrity, it's worth chancing. Today, as many games have grown more literal in their presentations, the minimum requirement for the player's imagination has diminished. However, for anyone responsible for crafting game content, the ability to imagine what's going on inside the player's head at any given time, and use this to spark the player's own imagination, continues to be key. Autoaim systems and other adaptive-difficulty tricks can make up for players' skill differentials, but I'm referring to the skill of interpreting a player's feelings beyond those of pleasure or frustration, which cannot be gauged as easily as how many health packs he's used or how many headshots he's scored. This is a skill that can only be cultivated from playing a lot of games, observing a lot of players playing games, or preferably both. And I think it's essential to crafting well paced experiences in a certain tone, especially if wit or humor is integral to that tone.

Think of a game that felt deeply personal to you. I bet you're either thinking of an incredibly time-consuming or competitive game with social elements like World of Warcraft or Counter-Strike, or the sort of game I'm writing about here -- the kind that recognized your expectations as a player and fed them back to you in a surprising, insightful way. The feeling you get from this is the videogame equivalent of meeting someone with the same favorite band as you.

Sometimes you're lucky to get this feeling early on in a game, such as in Half-Life 2, when in the opening scene, a professor is hammering you with exposition, leaving you free to explore his office... where you'll more than likely knock over one of his computer monitors, causing him to interrupt his speech and chew you out for not paying attention. It's a great little moment. The designers knew you were gonna screw around in there, and placed that little trap for you. In almost every shooter before Half-Life 2, you can bunnyhop like a moron in the middle of even the gravest of monologues. But in Half-Life 2, this opening scene informs you that you have a real presence in this world beyond just the presence of your guns. The rest of the game didn't have such moments in it but the placement of this very first one was important to quickly immersing the player in a world rich with detail, and setting an expectation for bits of comic relief amid all the sci-fi seriousness.

There are dozens of other examples, though on some level it's surprising there aren't more. In Uncharted 2, relatively early on there's a sequence in which you're introduced to the grenade-throwing mechanics. Nate Drake is well-positioned behind cover as a bunch of thugs bear down on him. Fortunately, a box of grenades happens to be in arm's reach. So of course Drake, speaking for the player, quips something like "well isn't this convenient", referring to the unlimited grenade box while acknowledging the cliché, subverting it, and turning it into one of the game's great self-referential moments. Without that little bit of nudging-and-winking dialogue, the scene would have been just another grenade-throwing tutorial.

Minor cases like that can be surprisingly poignant. In the classic SNES role-playing game Chrono Trigger, someone at some point makes the joke about how the protagonist doesn't talk much, a profound moment that for the first time points the spotlight at the silent-protagonist trope common to so many games. This same joke then appeared seemingly in almost every other game with a silent protagonist (including Half-Life 2), so it lost its impact after the umpteenth time, but the first few times it was great.

On the other hand, Saints Row does a brilliant job of subverting the silent protagonist cliché by having the protagonist blurt out some sort of ultra-offensive Silent Bob-style one-liner when you least expect it, just when the story has convinced you to start taking it seriously. It's legitimately funny stuff.

Then there's Max Payne, in which the hero once dreams he's a killer inside a videogame, in one of the game's most surreal and memorable moments. This too became a tired idea as more and more games fed you the line about how "this isn't a game", hoping you'd never heard it before. No More Heroes gets it right, by breaking the fourth wall on several occasions towards the end, conceding its own ridiculousness in moments that drive the game's self-effacing charm through the roof. 'This is a game' is No More Heroes' message, delivered playfully and perversely, at the height of the player's investment in the story.

Narrative recognition doesn't require breaking the fourth wall. Another one of my favorite examples is in The Elder Scrolls IV: Oblivion, in the excellent Dark Brotherhood quest line, where you've decided to become a contract killer and work among a surly bunch of assassins who don't take kindly to newcomers. Just as you finally grow to like them and earn their respect, you receive a secret mission to kill them all. It's timed perfectly. You kill scores of things in this game but none are tougher to put down than those guys in the assassin's guild, because the designers made you work for their affection, made you feel attached to them, and only then asked that you turn on them -- a move that's made less despicable knowing that you're all following the same strict code of conduct.

There are simpler and cleaner examples than this. In the original Halo, the option to invert the controls for looking up and down is couched as part of a diagnostics test at the beginning of the game, when you're learning about your special armor. A lesser game would have thrown this option into a menu and not integrated it into the narrative. But the makers of Halo figured correctly that the last thing a guy like me would expect was a fictional justification for inverting the Y-axis on my controller. All I expected was to be bored by a tutorial.

What do all these games have in common? They have in common these crafted moments designed to anticipate exactly what the audience is feeling about the game at a particular time, and feed back on those feelings, sometimes merely by acknowledging them. I know you're frustrated. I know you've done this before. I know you don't think you care about these characters. These types of concessions can have a real impact. Narrative reversals are integral to storytelling -- I'm referring specifically to when the game narrative is used to reference the act of playing, not the story itself. Such acts of recognition are at the center of effective game narrative. They're among the most meaningful ways for a game to communicate with a player. One of the best feelings in life is getting what you want without having to ask -- at a restaurant, at work, in a relationship, you name it. Games have the ability to provide this exhilarating feeling using narrative techniques.

Characters and gameworlds that tend to correctly guess at the player's feelings and feed back on them are the best kinds. They're typically bound to games that don't break when you push at their boundaries, as when you knock over that monitor in Half-Life 2, realizing you can knock over all sorts of shit throughout the entire game. Games like that let you explore the upper limits of the constraints of their rules. In last year's Demon's Souls, one of the first characters you meet in that oppressive and bleak gameworld is a cynical ghost of a man who's given up trying to escape. He more or less calls you a fool for attempting to do better. It's straight-up reverse psychology, and it's fantastic. Here's a game that knows it's challenging. Why shouldn't the world have characters in it who've become frustrated by those challenges? Not only does it fit the fiction, it affirms the player's own first impressions about the game. That ghost is a pretty sympathetic character. But then, if you want to kill him for being such a noob, you can go right ahead and do that too, says Demon's Souls. That game covers its bases, which is partly why its audience likes it so much. Those who gave it a chance tend to make a deep connection with it, because its designers knew you were going to have to fight for every inch in that game, and anticipated and fed back on the range of emotions you'd feel along the way.

Most games with a narrative component can and should have moments like the ones I've mentioned here, because their designers ought to be concerned with what gives their game its distinct tone, as well as the likeliest emotional dispositions with which their players will approach any given scenario. This happens when designers think of players as individuals rather than as groups. When games speak to you on a personal level, when you experience one of those rare but great moments of recognition and reflection while playing, that's not coming from some guy following up on data from a focus test or gunning for the sci-fi shooter market. Instead, it's coming from someone trying to reach you, with the confidence -- the presumptuous authorial arrogance -- that he knows what you must be thinking at a certain time. How dare he? And yet he really is speaking to you in that moment, through his game.

The Three -Mations

I consider a game's story to be the subset of narrative content authored for that game, which the player interprets to have some coherent meaning. The meaning does not necessarily need to be placed there purposely by the game's creators, it just needs to be perceived by the player through the intuitive pattern-matching we conduct as we go about trying to find meaning in all things. Story is just one of many types of feedback games can employ in response to the player's input. Thinking of story as another form of feedback, I think, is the key to tapping its potential in games -- as well as the key to integrating it into the design process in an appropriate way, while mitigating risk of the story derailing or constraining the design.

Game feedback, in general, provides information, confirmation, and affirmation. Story is no exception; I can draw an analogy between story and user interface design. In God of War III, a big spinning "O" button next to a prone enemy carries a specific, almost Pavlovian connotation: Here lies an enemy that you can savagely murder if you get close to it and press the corresponding button. The UI prompt is a reward, affirming your combat skills; it confirms the availability of a unique set of killing moves; and it informs you of the input required to access those moves. Good interface design in games is contextual and just feels right, in this fashion, because it's closely linked to player input and the affordances of the interface match the aesthetics of the game in general or the input in particular. Think of "You Are Dead", the game-over screen in Resident Evil. It works like a consolation prize, not a punishment. It's a pretty good screen.

Linear noninteractive game narrative tends to be ineffective because it's disconnected from the player's actions. However, there are numerous games with lots of noninteractive cutscenes that are still considered to have outstanding stories in them. These games range from Grim Fandango to Metal Gear Solid to Mass Effect. How is it that these examples break from the popular player-elitist idea that linear story is a bad thing for a game? The answer is that, in addition to having these lengthy story sequences in them, these games do the important small stuff too: inform, confirm, affirm through the story.

In Grim Fandango you can share your balloon animal shaped like Robert Frost with various characters, leading to unexpected comedy. In Metal Gear Solid, a character called Psycho Mantis looks for saved data for Castlevania: Symphony of the Night on your memory card, breaking the fourth wall as he points out your gaming preferences. In Mass Effect, you can shoot one of your party members dead if you don't like the way he's acting. By affirming the player's actions in these types of cases, these games inform the player of a wider-than-expected possibility space of interactions, which the player later confirms (or does not confirm) through additional interactions along the periphery of that space. These are examples of story working at its best. It's supported by a rich linear narrative in all three cases, which cushions these smaller, more-important story moments in deeper meaning. The result is three very different yet rewarding experiences that share in common a story-driven structure, where the player's input is acknowledged via story mechanisms and not just interface prompts as in God of War III. Nor are these games' stories simply playing out without respect for the player's actions, as is the case with many games reviled for their flashy noninteractive stories.

Animal Crossing is another excellent game that uses these types of techniques but does not have story in the traditional sense, because (following my first statement) it has no coherent meaning behind what happens in the story -- except for the thinnest possible premise of "Tom Nook gives you a home loan and eventually you pay him back." The game has the perfect amount of story for something whose underlying meaning concerns the simple pleasures of life. But even though Animal Crossing would never be considered a game story on par with Grim Fandango, or Metal Gear Solid, or Mass Effect, it is just as effective in its application of game-story principles. One of my favorite examples of this is in the game's clothes-making mechanic, where you can "paint" clothing patterns and put them on display in the local clothes store. This type of player expression is very common in games, which routinely let you change your character's hairstyle, paint your car, and so on. However, Animal Crossing takes these types of systems a step further by affirming your efforts, as when you find one of your town's animal residents wearing that shirt you made and complimenting you for your artwork.

Games rarely go to the trouble of confirming, let alone affirming, this type of player expression, as though we're all so confident in our ability to express ourselves that we don't want the world to judge or acknowledge our efforts. But, if you'll forgive another one of my analogies I've been meaning to commit to writing, think for a second about what happens the day after you get a new haircut. Either you're fishing for (or at the very least you're open to) compliments on your new look, or, if by some chance you're someone who knows her hair looks excellent, I figure you at least expect people (even just certain people) to notice the change. Conversely, you'd feel bad if no one noticed or seemed to notice. Doesn't mean you want to be fawned over in the way certain games vomit praise on you for your basest actions, like a mother indulging a spoiled child. You're sharp enough to observe subtle cues, as when someone who never once paid you a second glance does a quick double-take. And what a rush that is. Story in games should serve this end. All it really needs to do is acknowledge the player's input in a way that seems to give meaning to the player's actions, by being personal or specific or both. It also doesn't need to use a lot of words or cutscenes to accomplish this.