SAN FRANCISCO--On Wednesday night, Kim Swift and Erik Wolpaw were basking in the adulation of their peers, taking home top honors for innovation, game design, and game of the year in the 2007 Game Developers Choice Awards. On Friday afternoon, however, the two were forced to critically assess their contribution to the Orange Box compilation, detailing all the things that went right--and wrong--in a postmortem session.

　　"Don't hurt me! I love you!" As one would expect given the game's success, the presentation was jam-packed with attendees, who filled every seat and every bit of standing room in the hopes of gleaning some nugget of insight into Portal's development. Swift and Wolpaw provided nuggets aplenty in a half-hour lecture and a roughly 40-minute audience Q&A session. To start with, Wolpaw and Swift described their "delta theory" of integrating story and gameplay.

　　"By itself, our story wouldn't make a great novel," Swift said. "And the gameplay is all right, but a little on the dry side. Honestly, it would be a race to see which one would fail the fastest on its own. But because we had a really tight integration between the story and the gameplay, it really seemed to resonate with people."

　　The key then is to keep the difference between the "story story" and the "gameplay story" as small as possible, because wide disparities stick out to gamers. Wolpaw gave the 2001 shooter Clive Barker's Undying as an example of a game with big disparities. The gameplay consisted of a World War I hero frantically gunning his way through a haunted castle, with occasional pauses while he calmly interrogated levelheaded butlers, maids, and other servants about the horrors unleashed around them. According to Wolpaw, the team was ruthless about not allowing the "story story" to ever intrude on the "gameplay story" as it did in Undying.

　　Wolpaw also mentioned how the constraints the team was put under actually fostered their creativity. For most of the game, the computerized antagonist GLaDOS is present only as a guiding, encouraging, menacing, or taunting disembodied voice. While that built tension to the game's final boss battle unveiling of GLaDOS "in the flesh," Wolpaw said that presentation--and several other design elements--had more to do with the team's constraints than anything else. With only a handful of people working on Portal and a limited budget, the time and money demanded to implement other human characters (what with their animation and voice work) was unrealistic.

　　However, there was another purpose to having the player navigate the world of Portal utterly alone. Wolpaw had read some US Secret Service documents on interrogation techniques, and he discovered that when people are isolated for extended periods of time, they tend to develop affinities for inanimate objects. That was one reason the Weighted Companion Cube worked as a "character" in Portal, but not the only reason.

　　The cube, and its fiery end at the hands of the player, served as a perfect training sequence before the player headed into the final boss battle with GLaDOS, where various parts of the supercomputer needed to be tossed in an incinerator. The demise of the Cube made for a more satisfying level ending than simply making it through an obstacle course while carrying a random box, or leaving the box behind to get to the next obstacle course. Swift also said that players learn better when they're not stressed by incoming fire or a strict time limit, making the whenever-you're-ready forced murder of the Cube stick in players' minds better. Finally, the Cube's demise offers a pleasant symmetry of revenge to the final boss battle, which sees the player burn GLaDOS the way GLaDOS made the player burn the Cube.

　　Above all, the advice Wolpaw and Swift wanted the audience to carry away from the session was that developers need to embrace their restraints and treat them as fuel for creativity. They also need to have faith in themselves, their writing, and their team. Finally, play-test, play-test, play-test, and then play-test some more.

　　From the first week the development team had something to show, Portal was being play-tested regularly. It not only helped them find out what players wanted from the gameplay (and adjust accordingly), but it also helped them discover what players were feeling from the story part of the game, and work on better reinforcing those parts.