So apparently, believing that a ghost is watching you makes you less likely to cheat.
The study that determined this is actually kind of clever. They set up some difficult test, with a financial incentive to do well. They also make it possible to "cheat" on the test. (The software for the test was supposedly defective, and would sometimes show you the answer before the question. Test subjects were asked to press the space bar to clear the answer when this happened.) And some of the test subjects were told that some people claimed to have seen the ghost of a dead graduate student in the room.
Perhaps because they were scared of a dead graduate student's ghost hanging around, participants in the ghost story condition performed worse than those in the other two conditions on the mental rotation problems. More importantly, participants in the ghost story condition pressed the space bar faster than those in the other two conditions (though the difference between their times and those of the participants in the in memoriam condition only approached significance) when the answers were "mistakenly" shown prior to the problems. It appears, then, that thinking about the presence of a ghost in the room made the participants less likely to cheat by looking at the answers before getting the problems.