Found this through a BBC article, though here's a link to the University of Rochester's write-up, about a study (from the UK's Oxford Internet Institute* in associate with the University of Rochester in the US) that looked into whether they could make players aggressive in a game that they've removed the violence from but instead varied whether mechanics were explained. Basically they showed that the frustration caused by badly explained mechanics or just a high difficulty could cause an increase in aggression irrespective as to whether the game had violent content or not.
It probably seems like an obvious thing to many of us, but it's good to finally see a study into it (and cool to think they modded a game to do it - though it's interesting to point out the difference in emphasis between the BBC and Rochester's account on this).
I don't have access to the full study (neither having clearance to just see it, nor the money to pay to see it) so if any of you do, feel free to talk more about it (I'd like to know a bit more on how it was carried out). If you do want to try looking yourself, it was published in "The Journal of Personality and Social Psychology".
It should be noted that like most studies, it was taken on college undergraduate students (so keep that in mind).
*Sadly I couldn't find a link on the OII website, but I did find this upcoming event for next monday on how difficult it is to test aggression - maybe they are related (the event seems to be focusing on a standard process in aggression research that is miss-used, saying that testing aggression is a lot more complex that it seems - makes you wonder if their own study fell foul of it).