Mail from Harun Morrison a while back pointing to some notes on testing computer games, based on his experience doing just that at EA. I liked these in relation to The Broken World - esp around glitches, and the cross-connection of one thing to another, topics which fascinate me. There's nothing so good as the scent of accidental as it appears in a work (world) that is all intention to make me sit up and take notice.

..if you eat a yellow star you can fly for 10 seconds, [so] for the game to be satisfactory you shouldn’t be able to fly if you haven’t eaten a yellow star. There is also the physics of the game to be tested, is the length of time you hold a button proportional to how high you jump? Can you shoot yourself? Can you run beyond the borders of the game, can you stand outside the play space? Do the buildings have the integrity they should? Can you run through a wall that operates as the boundaries of the game, or does your avatar half-merge with something that should be solid? Do bullets pass through an enemy that is supposed to be susceptible? When the game is set on its most difficult setting is it too easy to complete, is the beginner setting to difficult for an average player. Do the multi-language formats work? Does the sniper rifle make a sound programmed for the machine gun? These are ‘bugs’, it is the aim of the testers to find them, be able to recreate them and communicate to a third party where and how the bug occurred...let the programmers (unseen on the floor above) know where the bugs are, and how they reveal themselves, the programmer would then fix the bug, when he or she felt this had been done, we would be sent an updated version of the game, we would try re-locate the bug or re-create the circumstances which led to its appearance.

What I also really liked in Harun's text was this stuff that starts to push at the difference between playing and testing:

Importantly the testing is executed by the same means the game would normally be played. We use the same control pads often the same button combinations. In fact the game could only be usefully tested this way if it is to successfully recreate situations in ‘real-life game play’. The process of playing the game and testing it are the same. However playing and testing are not the same. There is a difference in intention on the part of the operator which alters the definition and labeling of the activity from the point of view of the operator but not universally. An outside eye looking over the shoulder of the operator would not necessarily be able to define the activity. The outside eye might label what the operator considers to be testing as irrational or bad playing. If one thinks of a videogame as a system designed for an expected function, in this case playing, the misuse/ misapplication/ deconstruction/ destruction/ critique of the system is an activity that deviates from the expected function (defined by the designer). Perversely, testing is an activity that we communally recognize as operating by (the inclusion of) deviation from expected function and normative use/application.

More from Harun here.