Monday, March 23, 2009

The Refrigerator Box

Monday, March 23, 2009
In Oliver Sacks' The Man Who Mistook His Wife For A Hat, there is a chapter called "The Disembodied Woman" (excerpted here).
In it we read of Christina, a woman who has lost all sense of proprioception. She learns to puppeteer her own body; if she is not looking at a limb, she has no internal sense of what it was doing.

A nightmare scenario for anyone, but one that sounds oddly familiar to gamers.
Proprioception is what we aim to replicate, albeit distantly, when we talk about "body awareness" in games--to look down and see your body is the most basic feedback for a body we cannot sense.
I want to use this as a way to look more generally at the sense of immersion, of being embodied in the avatar, and how FPS games are limited in this respect. Before we talk about any design changes or solutions, we need to understand the problems.

We'll never be able to capture full proprioception (nor would we likely want to), but that same sense of "being there" is critical to immersion. Lack of proprioception is much of the difficulty for new players to grasp FPS controls, given the relative dearth of sensory feedback, and why even experienced players may become injured or killed in the world in a way that seems "unfair."

If FPS players ever feel as though they're a walking refrigerator box in the game instead of a person, it's because they are. Many games refine hit detection for the purpose of assigning damage, but collision-wise most player characters really are just ambulatory refrigerator boxes with a small view slit and a gun, and games are designed to suit.

This box-and-gun model has implications that have shaped and limited FPS games more than we might care to admit. First we'll talk about the player's field of view, then basic collision in the world and scaling issues, and finally a little bit of what this means for the game space and interactivity.

You may recall the furor over "true widescreen" shortly after Bioshock's release. Without getting into that fairly snarled topic of debate, it's worth noting that it was all over a few degrees of field of view. With the standing box-and-gun model for the player in FPS games, the player's field of view (FOV) is a critical window to the world outside the refrigerator box.
In reality we enjoy 140 degrees of binocular vision, where the overlap between the two eyes' vision gives us depth perception, as well as an additional 40 degrees of peripheral FOV.
Stepping inside the refrigerator box gives us monocular (i.e. non-depth perceptive) vision limited to about 90 degrees, and typically far lower, all with no periphery. (A move to third person would expand our practical FOV as well dramatically improve our body awareness, but with other far-ranging costs and implications to the game design, so we'll talk about that later.)
What makes this so troublesome is that maximizing player FOV in order to play the game effectively--situational awareness and target acquisition and so on--can run contrary to other design imperatives.

FOV 75 and FOV 90 on HL2: The Lost Coast. The man appears more distant at 90, even though we are standing face to face. Note that the crowbar remains the same, as it is rendered separately to prevent distortion.

A well known case is that of Half-Life 2. Evolving from the player's fondness for the non-player characters in HL1, Valve spent a great deal of time and effort modeling, animating, and voicing NPCs like Alyx and Barney, characters that would make up the emotional core of the experience.
They quickly found that their wide-angle FOV, ideal for fighting enemies in the world, made these friendly characters appear farther away then they would have liked, even when standing face to face (or face to box). HL2 shipped with a standard FOV set to 75 degrees instead of 90, with a special case that opened it up to 85 degrees during vehicle driving sequences.
Bethesda Softworks would take a different approach to the same problem for The Elder Scrolls IV:Oblivion and then later Fallout 3. They retained a higher, more "distancing" FOV, and chose to dynamically zoom in on any NPC in the world that the player talks to to simulate the desired visual intimacy.

In addition to the relative perceptual difficulties related to FOV and characters in the world, higher FOVs introduce distortion, given from projecting a spherical perspective onto a flat screen. While this can be cleverly incorporated into the game, as with playing the alien perspective Aliens Vs. Predator with a very high "fisheye" type FOV, it's a problem for weapons in first person.
And when you're a refrigerator box with a hole cut out, the gun you hold in your hands is a central visual hallmark, your primary (and often only) agency in the world.
At higher FOVs, the gun in hand begins to distort along with the world, its apparent distance from you potentially more disconcerting than the apparent distance to NPCs. That the gun is often rendered separately from the world may ameliorate this, but alienate the player in other ways, furthering his or her apparent separation from the world. They're already a firm proprioceptive barrier away from the game already; any additional layers between the player and the world become prohibitive.
But this is enough to sketch out the myriad of issues associated with the first person perspective and FOV; I'd like to talk about the box out in the gameworld.

People understand that their avatar is clumsier than themselves, and that with practice a player is able to overcome or compensate for most shortcomings. This means they're nominally past thinking about the refrigerator box, but the box is always there. Often we can feel it, more often if the gameworld has not been scrupulously designed to accomodate the problem.
What I mean is that with level design there's a tricky balance between visual scale and player scale. The immersed player forgets that he is a refrigerator box, yet the world still need be scaled to accomodate his generous girth.

A friend of mine and fellow level designer Drew Risch taught me the simple, fundamental rule for good playspace in an FPS: keep it wide.
It's relatively simple to establish scale for all buildings, objects, and characters in a game to correspond with reality, but real world scale is unbearable for a refrigerator box to negotiate, even if the box itself is the relative dimensions of a person. You're a box, a box with only a small hole out the front with which to engage the world, so allowances must be made. Despite being a box you must be made to feel the hero, and heroes don't have to carefully negotiate kitchen furniture like the morbidly obese.
To that end the gameworld must be carefully widened, ballooned to allow the passage of this refrigerator box player and his often as-clumsy-if-not-clumsier AI compatriots.
The trick is to make this work mechanically, while still visually keeping the world in the ballpark of reality. It won't match, not if the world is truly wide enough for the player to move around comfortably, but these discrepancies can be carefully masked. In HL2, the corners of the Gordon-box can get the player stuck neatly into right angles of tight spaces such as elevators--level designers place invisible wedges in corners to prevent exactly this.
(In third person these sorts of issues are even more pronounced, though in different form; replay the beginning of Max Payne 2 and observe a world necessarily warped to accomodate Max's bullet-dodging antics and follow camera.)

We begin to see clearly why information systems, usually delivered via some sort of heads-up display (HUD) become so critical to helping the player over such considerable handicaps. The player has a narrow FOV and no proprioception, so environmental cues of sight and sound (plus a little rumble, if on a console) are all he or she is going to get.
Detailed sound design with corresponding sensitization of the player that these sound cues mean something (like the dry snapping sound of a bullet passing close by) can help close to gap left by so many missing senses, but won't fix everything. Which is why damage indicators and the like are now considered standard, as there is simply too much critical feedback the player needs to survive.

But I feel as though we've opened many doors without much sense to how we might approach a solution or a change from the problems these issues present.
At this point, what I want to impress upon you is that the player character is a refrigerator box with a hole cut in it and a gun sticking out, and so long as he remains such--a mute, verb-poor box, murdering everything in his path only to pause occasionally to flail at an oversized green-lit button or watch a mandatory cutscene--FPS games will likely remain the dim spectacle that they've been for years. There certainly is hope, but there's also a lot of dead weight carried from sequel to sequel that ought to be reconsidered.
But I'm getting ahead of myself and betraying more ambition than I care to for a first entry.

Next time, we'll talk about methods to overcome these seemingly crippling limitations, and explore other design issues present with the status quo.

Summary of embedded links:

The Man Who Mistook His Wife For A Hat on
Excerpt from The Man Who Mistook His Wife For A Hat, "The Disembodied Lady."
Wikipedia entry for Proprioception
Wikipedia entry for field of view
Valve Developer Wiki on FOV in HL2
Excerpt of Max Payne 2 playthrough on youtube


jeffdr said...

Right on with the FOV analysis. I'll add a bit more from the engineer's perspective.

A big part of the problem is the display. Picture your self sitting in front of your TV or computer monitor. Now with your "mind pencil" mark the angle your screen subtends. What is it? For most arrangements it's usually somewhere in the neighborhood of 20-30 degrees. See all that stuff that's not your TV screen? That's what you're missing in games today. Even giving the player a "narrow" FOV like 60 or 70 is actually being pretty generous and erring a lot on the side of distortion.

Also its a subtle point but the way perspective projection works in real time rendering is not even as good as a "fish-eye" style camera lens. As straight lines cannot be bent in realtime projections the distortion is actually even more irregular.

As a guy who's worked in the VR field and as an owner of ridiculously large monitors, I'm already pretty sold on peripheral vision.


PS - Figure 2 is going directly into my internets folder. Nice work

Johnnyburn said...

Looking forward to the next installment, gauss! The refrigerator-box picture is an instant classic.

Jeff - Although I had to look up "subtend" to be sure, that is a great point about straight-lines in 3D projections.

It is interesting to me that the narrow field of view makes the player move their view around so much, which forces them to strafe. Although I don't walk sideways very much in real-life, I think that the strafing technique captures the sense of disorientation of not looking where you are going.

That sense of disorientation is definitely part of the immersiveness mentioned above.

For some reason, the method of strafing that has evolved is more intuitive to use than the "tank and turret" or Mechwarrior type of control. It makes sense to think of the body with a pivoting head on it (tank-and-turret), but it is easier to control a head with pivoting body underneath it (strafe).

-john b

Jack said...

Jeff--excellent point. Did you check out the Valve dev wiki link for FOV? There's good information there, especially pointing out that if you want to properly present wide-angle FOVs, you're going to need a curved screen (like an IMAX).
Was not aware that we're not even getting "accurate" distortion. Interesting.
The VR field angle is something I forget about... I think particularly with FPS and the refrigerator box problem there's a lot of lessons to be gained. Looking forward to comments from you in the future.

John--thanks man. And there are a whole host of issues that come with movement speed, among them the tendency to strafe, as well as whip the mouse around constantly to do threat assessments where you're missing peripheral vision.
(As a side note I continue to be amazed at just how incredible our own visual system is--we don't get all that readable picture out of our periphery, but we get better-than-center night vision and excellent motion detection.)

I like that you mention mechwarrior/"tank" controls, because playing RE5 is one of the few games that has the player move and shoot somewhat more in this fashion, albeit in separate moving and shooting stages. Alternate control schemes are appealing to me, even for FPS games, because a control scheme is ultimately how you describe the most basic forms the way the player interacts with the world. People grouse about the "tank" controls in RE5, despite the addition of strafing, but what would be an improvement? Making the characters control as they would in Halo? Why ruin the game's unique character?


Ninjas said...

I really like high FOV more, but the perspective has always bothered me a little. I would like to see, at least as a test, a full screen fish-eye distortion shader at some point. I think it would probably look pretty cool-- here is the raytraced ET:QW demo intel did awhile back:

Nezuji said...

Heh, a little late to the table here, but Ninjas, you might want to check out PanQuake. I never got to try it out or see it in motion, but it's a 360-degree FOV adaptation of the original Quake engine. That is to say, the very edges of the screen actually represent the view directly behind the player character.

gauss said...

Don't worry, he still comments--and he's a friend of mine, so I'll make sure he sees this. Thanks!

Broklynite said...

I never really gave it much thought. I've fiddled around with FOV controls and always set them to maximum- it seemed to make sense for being able to see enemies coming at me from the sides. I never noticed any distortions. Then again, I never really looked for them. Nor did I ever feel a distance problem from people I went face-to-face with. But then, I don't usually go face to face with anyone unless my shotgun is out. I personally dislike the zooming-in-on-the-face used in Fallout and Oblivion- especially when it glitches and you end up talking to someone halfway across the map. It really jerks me out of the suspension of disbelief. Mind you, I enjoy both games otherwise, so I don't mind it too much.

Okay, that being said and out of the way, I'd be curious to hear thoughts regarding home theatre projectors. I gave up monitors a few years ago, getting a 720p LCD projector which is hooked up to my computer, my xbox, my cable, everything. It spoils you, I must say. But that said, I would be curious about how that might affect the view. Rather than taking into consideration a 15-20" monitor sitting 2 feet from your face, what are the effects of having a 110" screen ten feet from you? Can anything be done with FOV to take that into consideration? If so, what? And how would it affect everything. Projector owners are still in the minority, but they are a fast-increasing percentage. I would be curious to see an option in games for projectors. But as I said, I'm curious to hear from professionals you opinion about things that could be done. Keep in mind, a projector will mostly fill your actual vision (well, okay, not mostly- but a good deal of it).

gauss said...

Broklynite: thanks for the comments. Different displays don't really change many of the issues cited here, unless it's a multiple monitor solution extending your peripheral vision to either side. But I suppose a home theather setup would be good in a similar way, because depending on where you're sitting, it might cover most of your actual field of vision.

If this is the case I would fish around for some solutions to widen up the FOV and see how it feels when you play the game.

Dhatz said...

it's the metod of rendering that prevents form accurately sensing the image on large screens,as it gets distorted because it ignores screen size and viewer's distance from screen.

Lord DoomRater said...

Grab's latest version to see how this problem applies similarly to boxing games and other fighting games that rely on hitboxes. Here, it's not so much a distortion of vision (in fact, the side angle tends to be BETTER for traditional arcade boxing games, though seemingly less for this one due to the realistic physics actually playing a part) but the fact that despite the good looking skins, the game really marginalizes the boxer's hit detection as simple shapes. It's not quite as extreme as a single fridgebox, but try punching someone in the face in that game and you'll see what I mean. It looks like the shot should be doing damage but it doesn't always do that- and the punch is maybe a few pixels off into the gloves or a few pixels off the face. In reality if someone's hands were down it would be a cinch for someone to throw a proper punch into another man's jaw, and that's despite the fact that our FOV actually forces us to first judge the distance between our hand and the jaw!

Anonymous said...

Ha, figure 2 looks like the Daleks and Snake had a love child.

Post a Comment

gausswerks: design reboot. Design by Pocket