The Well-Made (Video Game) Plot, Part 2

Continued from Part 1.  You may also want to read Belinkie’s post on the same subject and this older post by Perich.

Narrative Exploration vs. Narrative Discovery

Crucially, I can’t discover anything that I know will be there.  (This is one of the reasons why we talk about explorers “discovering” places where other people have been living for generations, although ALL of the other reasons are straight-up racism.)  Finding Xanax in my friend’s medicine cabinet would usually constitute a discovery, but not if they have “Xanax 4 Life” tattooed across their navel.  By the same token, I can’t explore a space I already know well.  I can discover things in my own apartment if I forget that I have them, or if someone plants them there — like, if there’s a surprise cake waiting in my refrigerator, for instance — but I can’t really explore that space, because I already have a detailed  (though faulty, because cake-deficient), mental map of its layout and contents.  This common focus on the unknown is why we think of exploration and discovery in the same breath.  There are situations where I can discover, but can’t explore (the cake scenario), and situations where I can explore, but can’t discover (because there’s nothing out there), but exploration is often a precondition for discovery.  If I don’t go wandering around in the woods, I can’t stumble upon the clearing with all of the wildflowers.

Well-made plots hinge on discovery.   Learning the murderer’s identity in a mystery is (or should be) a big moment of discovery, as is the point where we realize the crucial facts about the serious dramatic protagonist’s psychological makeup, or what have you.  But a huge, huge part of actually reading fiction is about exploration. Not every detail of the plot is actually critical to the plot.  Few are, in fact.  Even in a detective story, which tends to be weighted heavily towards discovery, there’s an aspect of exploration as well:  the main character explores the seedy underworld, and then discovers who committed the crime.

In fiction, we need to put the big moment of discovery at the end, even though in the real world it might not work that way.  In real life, the detective might put together a strong circumstantial case early on, spend a few weeks more looking for the smoking gun, find nothing, and see the crook convicted on the strength of the circumstantial case anyway.  In narrative terms, that’s not satisfying.  There can be a steady stream of discoveries throughout the novel, but as you get to the end they need to be bigger and thicker on the ground.  By the same token, a story’s beginning will tolerate large swathes of directionless exploration in a way that the end simply will not.  And it’s not just detective stories where this matters.  If I actually am in therapy, I will probably continue to explore my psyche for months or years (or even indefinitely) after I discover the big issues shape my interaction with the world.  But who wants to read about that?  So in serious works as well as genre fiction, the beginning of the narrative tends to be weighted towards exploration, and the end towards discovery.  Note that in a classic whodunnit, ruling out a suspect is a major point of discovery — and so the curve I mentioned in Part 1, where you eliminate suspects faster and faster as you move towards the end of the novel, automagically gives you a structure with lots of exploration at the beginning and lots of discovery at the end.

So both detective stories and “standard” narratives depend on the combination of exploration and discovery.  And in both of these, we want exploration and discovery to be teleologically linked.  We want to feel that the important discoveries arise organically out of the protagonist’s process of exploration.  We want to feel that every wrong decision is bringing us one step closer to the right decision, right from the very beginning.  Otherwise when the solution finally does appear, it feels like deus ex machina.

On the other hand, Whendunnits (flashback-structured narratives where we know something will happen, and we’re just waiting for the hammer to drop), do not really tend to feature exploration and discovery in an important way, at least not within the narrative.  This is because the characters are unaware of the plot’s central mystery.  They can’t be trying to figure it out:  they can’t find the Xanax because they aren’t even aware of the medicine cabinet.  Here, exploration and discovery are things that happen to the reader, not the characters.  We feel that we are “exploring” the book (or film, or whatever) when we are not encountering the Big Important Event that we are waiting for, and “discovering” something when we finally encounter them.  But in this case there is no causal link.  The hundred pages of descriptive prose come before the critical action scene not because one leads to the other, but because we read books in order from start to finish.  Furthermore, the whendunnit depends in a very important way on the dead-end false start.  Before the car actually comes crashing through the window, it needs to make several premature attempts that fail.  You need narrative threads that seem like they’re going to lead to the big fated event, but which actually go nowhere.  They’ve got to go nowhere — reserving the important events to the end is how whendunnits build tension!  Which means that our “exploration” of these false threads is only mechanically connected to our eventual “discovery” of the real one.   Kill Bill is a handy example here.  Whenever the Bride squares off against one of the other Deadly Vipers, we know full well that she’s going to end up killing them.  Tension in this kind of showdown (like in other kinds of showdowns) depends on our not knowing when the fatal blow will be struck.  Every time the Bride takes a swing, then, we’re seeing one of those dead-end false starts that I was talking about.  These are important, and exciting, and in a very real way what the fight scenes are all about… but they aren’t necessary. In retrospect, they don’t contribute anything to the swing that finally does connect.  (Movie fights are in this regard quite different from real life fights, where wearing down your opponent is a real and important strategy.)

Of course, Kill Bill isn’t actually structured like a flashback.  We know how that plot will turn out not because we’ve been told, but because of the conventions of the genre.  As I mentioned in Part 1, almost any story has a meta-whendunnit aspect if we know the genre well enough.  If we know how things tend to work out in this kind of story, we’re going to be wondering when they will work out in that way that they do.  Interestingly enough, the whendunnit’s charactaristic separation of exploration and discovery appears here too.  It’s important, within the world of the story, for the detective’s exploration to lead to the big discovery that solves the murder.  But our external sense that a murder-solving revelation of some kind is “necessary” does not have an organic aspect.  It’s just how stories like that work.  They could work another way, after all, many stories do.

Exploration and discovery in video games

I’m super, super glad that text adventure games came up in the comment thread to Belinkie’s post.  Obviously, playing these games requires both exploration and discovery.  There’s a world to wander and explore.  There are also things to discover in it — most obviously the objects, which can be picked up, manipulated, and collected for points, but also discrete pieces of information about the game world.  So if I play King’s Quest III, for instance, “exploration” involves visiting every square on the world map, “discovery” involves finding the secret lever on the bookshelf (which in turn opens up a new wing of the Wizard’s house for exploration), or getting the porridge, or learning that you can crumble the cookie up in the porridge, or learning the desert on the west edge of the map simply stretches on forever and will kill you if you wander around in it for too long.   Neither one of these, alone, would make for a very satisfying game.  The flash game Doodle God presents you with a “discoveries only” model:  no world to explore, just a bunch of objects you can try to get to interact.  It’s kind of fun for a while, but it gets real old real fast.  For an “exploration only” game, well, may I direct you to Google Maps? Note that even these examples aren’t totally pure.  There is still something to explore in Doodle God — trying to exhaust all the possible combinations is more or less equivalent to visiting every square on the world map — and the good people at Google did slip an easter egg or two into the street view images.  But the exploration in Doodle God is faulty because the explorable space is boring, and the discovery in Google Maps is faulty because the ratio of goodies to dross is unacceptably low.  And both are faulty because they don’t relate the discovery to the exploration.   Exploring possibility-space in Doodle God only leads you closer to your next discovery in that you eventually run out of wrong answers — exploration and discovery are mechanically linked, but not organically.  And nobody finds the easter eggs in Google Maps by consciously looking for them. Either you find them completely by accident while trying to get directions, or you read about it on the internet, right?  From this, we can derive a couple of narrative laws of videogames:

•  The explorable space needs to be interesting.  Beautiful, funny, cute, ugly, all of the above, just don’t let it be boring.  Of course, you can’t make everything equally beautiful, or funny, or cute, or even ugly, because all of those will get boring eventually if you’re bombarded with them…

•  The ratio of goodies to dross needs to be pretty high.  Shadow of the Colossus, which contains only a handful of things and less information, probably represents something like a hard limit on this front. It’s a beautiful game, but you couldn’t get away with a game that had much less in it. (And of course to a very great degree, Shadow of the Colossus works because it’s an exceptional case.  If all games were that spare, it would get old.)  Of course the ratio can’t be too high either, or the goodies will cease to be goodies…

•  Exploration and discovery need to be somehow linked.  This deserves some further unpacking.

One thing that separates video games from other kinds of narratives is that the player can partially control the pace of events.  In interactive fiction and its related genres, you usually have the choice of 1) moving on through the game, i.e. taking actions, advancing the narrative, and 2) lingering on the description.  In normal, non-interactive fiction, the author takes care of this for you.  Tolkien’s The Hobbit begins with an extended description of Bilbo’s Hobbit-Hole, and Bilbo himself, and Hobbits in general, which runs on for four pages before Gandalf walks over the hill and begins the plot proper.  This descriptive chunk includes the sentence “It had a perfect round door like a porthole, painted green, with a shiny yellow brass knob in the exact middle.” Now, if this were interactive fiction, you wouldn’t have to let it go at that.  You could, for instance,

X DOORKNOB

upon which you would be given a slightly more detailed description of the doorknob.  You could also X the door itself.  You could feel the doorknob, kick the door, knock and see if anything happens, walk down The Hill to Bilbo’s neighbor’s house and see what that doorknob looks like, and so on.  Any aspect of the environment can be explored in as much detail as you like, or at least as much detail as the programmers bothered to add.  None of this will advance the plot.  In fact, unless the author is feeling particularly ambitious (mobile NPCs are tricky — although the real Hobbit IF game did feature them, as it turns out), the plot will be frozen in place forever until the player either accidentally or intentionally does something to start it, which is to say that Gandalf won’t walk over the hill until you’ve taken some predefined action to trigger the sequence.  Ah, but what action?  If you have a copy of the novel handy, you’ll probably come up with the idea of lighting your pipe and blowing a smoke ring, but what if you don’t have the novel?  In that case, smoking the pipe to trigger Gandalf’s arrival is something you’re going to have to discover.  And this is potentially a problem.  After all, it’s not like smoking a pipe is generally a good way to summon wizards, or indeed to accomplish anything at all!  And it’s not an action that is likely to arise organically out of something like examining a doorknob.

This is arguably a design flaw.  The player is meant to be controlling Bilbo, not Gandalf.  At this point in the story, Bilbo is passive.  It’s very, very important to his larger character arc that he has not done anything to trigger or deserve Gandalf’s approach.  But we have to distinguish between Bilbo’s desire to do nothing and the player’s desire to progress further in the plot.  Only the player’s desire matters, in this case.  So although the game will make it seem like Bilbo is the one who is actually doing whatever we ask him to do, the exploration and discovery involved here are player’s, not the character’s.

This means that the game needs to communicate to the player, giving them hints on how to succeed.  The discovery that you need to smoke your pipe is unlikely to arise organically out of standard IF exploration commands such as X DOORKNOB unless the programmer has accounted for this possibility, which if they’re a good programmer they will have.  The description of the doorknob after all can be anything you want.  It could be, for instance, “Smoke your pipe to summon the wizard!”  Crude, perhaps, but effective.  Better is something which communicates this indirectly, maintaining the fiction that the game is really talking to Bilbo.  “You don’t have time to look at doorknobs right now:  your morning nicotine craving is upon you!”  Even that’s a little obvious, although this early in the game, and with this arbitrary a puzzle, obvious might be fine. But you certainly don’t want every wrong answer to spell out the right answer for the player.  So it’s far more common just to use the description to indicate that the line of investigation has yielded all that it is going to yield.  If the programmer is in a thorough and snarky mood, X DOORKNOB might trigger the response:

“It is shiny, yellow, and made of brass. It is in the exact middle of the door.  I don’t know what else you were expecting.”

More likely, and more boring, is the standard brush-off text:

“You don’t notice anything interesting about the doorknob.”

Or worse still:

“I don’t understand DOORKNOB,” indicating that the lazy programmer hasn’t even bothered to train the parser to recognize “doorknob” as a noun.

All of these, though, have the effect of driving the player away from the fruitless examination of the door and closer to the right answer of smoking the pipe.  This is subtly different from the Doodle God example.  There, putting in a wrong answer only eliminates a single answer.  Here, we’ve eliminated a whole class of different doorknob-related actions:  we could have tried turning it, rubbing it, pulling it, taking it (seasoned interactive fiction players try to take everything), licking it, and so on, but none of these are worth bothering with if the programmer didn’t even put in an interesting description.

Of course, we could suppose that our programmer is a sadist.  In this case, the useless doorknob will have a very detailed description.  Every verb that the player can think of involving the doorknob will yield a meaningful response.  It will be possible to go back into Bilbo’s Hobbit-Hole, enter his study, and read Fuzznavel Brandybuck’s twelve-volume Encyclopedia Doorknobbica for further information.  This kind of red herring can be extended as long as the programmer likes, providing he/she is a bastard.

The inventory system is another way of communicating to the player.  One of the first things you do, playing IF, is to check what the character is carrying.  In our little imaginary game, Bilbo starts out with a pipe, tobacco, matches, and nothing else.  This might as well be a road flare:  after looking around a bit and maybe trying to take the doorknob, the door, and the Hill, most players will naturally try to smoke the pipe.  They can’t expect it to summon the wizard (or more accurately, to trigger subroutine GANDALF|SMOKE), but they can fairly expect it to do something.  After all, why else would it be possible to carry it?  Note again the possibility for confusing the player, either unintentionally or sadistically.  What if the player starts with a whole knapsack full of gewgaws, only one of which is useful?  Or what if it is possible to take the doorknob?  The poor sap will surely never put it down again.  “There must be a use for this doorknob,” the player thinks, as they dodge Orcs beneath the Misty Mountains.  “Any day now with the doorknob,” they think, as they stab spiders in Mirkwood.  “Please oh please don’t let me drop my doorknob!” they moan, as they cling to the barrel on the River Running.  And then — oh God, those poor fools — when they’re on the Lonely Mountain, and they run into a door for the first time since the start of the game, they will immediately fixate on the idea that they’ve finally figured out what the doorknob is good for, and obsessively try to figure out a way to attach it to the cliff face so that they can pull it open (not that that makes any sense), ignoring the thrush until it dies of old age.

Constraint, in IF as in literature, is really important.  One of the cardinal sins of the genre is the “guess-the-verb” puzzle.  You don’t want your player to get stuck halfway through the game because they tried to FLIP the switch instead of THROWING it.  One of the holy grails of IF, therefore, is a parser that thinks like a person does, i.e. one that recognizes the words FLIP and THROW as synonymous with regard to switches.  Typing “flip switch” and getting the response “I DON’T KNOW HOW TO SWITCH THE FLIP” is annoying, because an idiot child would know what you were really trying to do. Similarly, it is really, really bad form to mention an object in your description of a room, but then tell the player that the object is not present when they try to interact with it.  Don’t tell me that “I don’t see any phone booth here” when you just went on about the phone booth at length.  Don’t tell me that “I don’t know how to push the phone booth” if I tell you to push it:  pushing is not something you have to figure out how to do, stupid machine! These complaints are valid.  Both experiences can be intensely frustrating, and besides, if you only program the parser to recognize the input strings that will move you forward through the game, you’d wind up with a terrible piece of IF (and arguably a piece of F which is not properly I). But there’s an upward limit as well as a lower one.  A parser that could actually understand the entire English language, and would let you perform any physically possible action, would make for a horrible, confusing, borderline unplayable game.  The limits of IF are necessary, because they let you know when you’ve discovered what you need to do to move on in the game.  And this brings up a very important point:  regardless of what the game’s plot pretends to be about, the player can only want to progress further in the plot.  An IF game is like a series of locked rooms, where the goal in each room is to unlock the door into the next room.  Of course a room here is a metaphorical construct, quite distinct from the rooms you move through while actually playing the game.  It might be a literal room, or the emotional state of an NPC, or the character’s own emotional state, or a certain amount of money in the bank account… the point here is that you move around between these stations as if they were points on a map, once you unlock the appropriate door.  And unlocking the door means discovery, of course.  You want to discover the right sequence of events to move you further forward in the game.  And this is what fundamentally separates videogame storytelling from all non-interactive forms.  Relatively few videogames are murder mysteries, but all are narrative mysteries (if they have narratives):  perhaps the character is trying to save the Crystals from the forces of Chaos, or to destroy an experimental piece of power armor, but what the player is doing, in each and every case, is trying to reverse engineer the right sequence of events to move you further forward in the game.  And although there are some discoveries that have nothing to do with gameplay, most of the discoveries that the player makes will be of the “what to do next” variety.  Video game plots are usually dunnwhats, in that you know who the character is but need to figure out what to make them do.  However, unlike literary dunnwhats, the choices for potential action are rigorously constrained.  In this, it’s more like the basic model of a whodunnit, where you choose from a handful of potential solutions.

This has some interesting implications.

•  Discoveries in games are not neatly distributed in time the way that they are in normal fiction.  They can’t be, because you never know how long it’s going to take the player to figure out what to do next.  However, a good game will often make some kind of effort on this front.  Players have less patience for wandering around searching for clues when they’ve been playing for twenty hours already.  Take King’s Quest III again, once you get to the endgame in Daventry, the game’s map becomes almost completely linear and you’re already carrying all of the objects that you need to solve the remaining puzzles.  Much, much less exploration is required.

•  There is no one “big mystery” that drives a game forward from beginning to end.  Even if your game is a murder mystery, deducing the culprit’s identity is going to be a puzzle like any other — more elaborate and more interesting, possibly, but still just a lock that opens a door.   When you solve the kind of arbitrary puzzle that makes up the bulk of all IF games, such as, say, building a makeshift telescope out of a paper towel tube and a pair of thick glasses, you’re not thinking “Yes!  One step closer to taking down the whole Barksdale organization!”  You’re thinking, “Yay, I solved that puzzle!  What’s the next one?” (By the way, The Wire:  The Text Adventure would be off the hook. “It is pitch black.  You are likely to be robbed by Omar.”)

•  Barring a weird, meta example, the character in this kind of game will never care about the questions the player cares about.

But so far, we’ve been talking about these games in terms of discovery.  Just like normal fiction, video games balance this with exploration.

If you track back up to the various “discouraging” responses I invented to X DOORKNOB, you’ll notice that only the “snarky but thorough” response is effective from an exploratory point of view.  It’s clearly wrong, and lets you know that, but it’s fun to read anyway.  Ideally, you want all wrong actions to have a response like this.  Trying to take the doorknob and getting a response along the lines of “You can’t get ye doorknob” is a failure of the game’s exploratory mode… although a forgivable one, because really:  a doorknob?  Still, getting something like “The doorknob is screwed tightly into the wood of the door, and besides, you don’t feel like carrying it around” is much, much better.  Sometimes, these error messages are so much fun that players will actively try to trigger them, exploring purely for exploring’s sake.  (The same thing is going on in games like Out of This World and Dead Space when players try to collect the whole set of lovingly rendered death sequences.) But the main function of these messages is to soften the blow of getting the wrong answer.  Which is to say:  exploration, in games, is mainly a function of trying to discover something and failing.

And we’ll get into the implications of that more in the third and final part.