I'd be interested in seeing studies that look at this sort of thing for a variety of different software types. My completely-unsupported-by-evidence hunch is th

I'd be interested in seeing studies that look at this sort of thing for a variety of different software types. My completely-unsupported-by-evidence hunch is that different lengths and types of software project do better or worse in silent environments (although I also suspect that the largest variance is from personal preference). By way of personal anecdote, when I'm working in the games industry, coding teams tend to be very small and very cross disciplined, and the smaller the teams the more frequently code changes touch other devs because you're often changing several systems across the engine throughout the day. In this environment myself (and co-workers - I did ask :) find non-silent, "noisy" environments work well. Incidental conversations and cursing are actually quite good ways to add interesting fluff to the games as they are being created, and cursing is a damn good barometer indicating that some of the APIs for my code is rubbish to work with. Conversely, I've worked on more traditional things where I'm deeply involved in a rabbit hole of my own burrowing, and in those cases less noise does seem to help. In those cases I'm general solving very narrow and specific problems, and the APIs I'm offering to other devs (if there are APIs) are far more formally defined. However, I'm likely biased. Unless it's a very specific and very hard problem I'm trying to solve, I like some level of background noise to keep the useless parts of my brain occupied while the code bits do their stuff. The more boiler plate the code, the louder I like the noise so that parts of my mind don't get bored and drift off into more interesting thoughts.