Zen and the Art of Level Design
11/19/97by erik robson
As I’ve picked up the pace with my own level design over the past few months, I’ve also started playing other designers’ levels with a more critical eye. Some very interesting patterns start to emerge when you sample a cross-section of the single-player Quake levels out there, and these patterns have forced me to think about my own designs in a different way.
There seem to be two distinct approaches to putting a level together:
For lack of better terms, I’ll call the first “narrative” and the second “abstract”. A narrative level will generally attempt to recreate a setting as realistically as possible, and take the player from a logical entry to a logical exit while adhering to a traditional narrative structure (conflict->climax->resolution). I consider Steve Rescoe’s work to be narrative in nature – we start off on a cliff outside the village, have to break in, flip some switches, defend ourselves, assault the castle, and make our way to the heavily-guarded exit. The layout, texture and monster choices, and puzzles have what fiction writers call “verisimilitude”. It’s like realism, but within a context that may or may not be realistic unto itself. You can make a movie about dinosaurs that have been genetically resurrected, but the people had better talk like people talk and the dinosaurs had better move like we’re used to seeing large animals move, or you’re going to lose viewers to disbelief. We can accept a fantastical premise as long as it’s anchored in reality, and a narrative level attempts to adhere to reality in as many ways as possible.
On the other pole is “abstract” level design. There’s no such need to cling to a realistic representation of a setting – nail traps line the otherwise utilitarian hallways of a castle; fiends are locked in closets that open only after a series of buttons have been pressed; lava and slime flow freely inside military bases. Many of the “Sadlark” maps by Jonas Lindstrom contain elements that I would attribute to abstract maps. (Note that I’m a big fan of the Sadlark series, and single them out only because they’re among the best levels in this style.) Rather than emulating a specific, recognizable setting, a map seems to have been consciously *designed* to antagonize a player. Monsters exist in this place only to attack. Cleverly-designed traps are built into the architecture, and one can almost see some insidious architect standing over his blueprints, scheming: “Heh heh… he’ll *never* make it past this room…”
Of course, it seems silly to envision these trap-ridden places as having been built solely to destroy a single space marine, but this reveals an important point. I’d argue that when you’re playing an “abstract” level, you’re not really in illusionary space to the same degree that you are in a “narrative” level. While E2M1 emulates a (fairly) realistic military base, with enforcers and grunts patrolling the halls and guarding their expensive electronics equipment, E2M5 is an odd mish-mosh of monsters, living in harmony in a medieval castle, their sole purpose in life to wait on their spot until a space marine comes into view.
I obviously have a personal preference as a player and level designer: Narrative maps strike me as more immersive and more logical in their decision-making. However, it didn’t seem right to criticize without first trying, so when I first started working these ideas out, I decided to make one of each and pay close attention to the methods involved with each.
The narrative level was to be a sequel to my Waterworks map. The new level would take place in an abandoned electric company building, complete with decomposing architecture and half-functional doors and lifts. The abstract level was to be a free-for-all: The architecture didn’t have to imply any specific use, and could be ornate simply to be visually interesting. It had only to be consistent. The architecture and layout were to revolve around the puzzles, traps, and ambushes awaiting the player.
I immediately found the narrative level more difficult to work with. The same formal rules of ambush and puzzle applied to this realistic setting, but there was an additional layer that needed to be satisfied: Does this puzzle make sense within this context? The raw puzzle won’t do; it needs to be built out of, or housed within, the trappings of the context. I wanted the player to perform some task on the first floor before progressing to the second floor. The task should make sense within the context of the abandoned electric company – it ended up being a light-based “puzzle”, similar to one in Waterworks, where the player enters a darkened section, and must re-investigate the previous area to find a way to turn the lights on. It’s not an Infocom-caliber puzzle, but it functions as a realistic problem that needs solving without calling too much attention to the fact that it’s a puzzle.
I must admit the abstract level was refreshing and liberating to work on, not requiring the painstaking emulation that the other map was demanding. I felt a bit like I do when working on an abstract painting – like I’m being selfish in my process, making decisions not with the viewer in mind but based on how I “feel”.
The immediate freedom that this rendered, however, may have benefited the level. An analogy is the difference in process between painting and sculpture – a single decision may take minutes or hours to execute in a painting, while a single decision can take days or even weeks to execute in a sculpture, because the *medium* tends to be slower. Similarly, a trap idea in a narrative level might require a number of architectural changes and/or additions just to make it work within the context. An abstract map will accept the trap in its “raw” form, and as such, the overall process becomes faster and more immediate.
When all was said and done, the narrative level felt more rewarding to have finished. The abstract was easier (and sometimes more fun) to work on, but in the end felt more like a formal exercise than a dungeon or arena. The Electric Company allowed for a greater depth of story (in Quake terms :) and a richer set of possibilities for atmosphere. I think that, in the long run, narrative-oriented, representational maps will be the key to using 3d engines for role-playing and adventure-based games. The levels featured in the Q2test are very narrative in nature, and I am hopeful that Quake 2 will set a better precedent for maps occuring within a context than Quake did.
The two levels designed in tandem with this essay can be downloaded at: