Well, I didn't say it would be easy to come up with such a design, only that it would be much more compelling if someone were to do so and then still get positive results. Statistics can be messy, and you can often come to completely different conclusions using the exact same data depending on how you actually select your computational set and how you do the math (this also ties in to problems with meta-analyses that have been discussed in other threads on the board).
The simplest way I can think of would be to only run each participant through a single trial, without warning them of the upcoming response-evoking stimuli. For example, tell the participants that you are simply measuring how relaxing classical music affects skin conductance, and put a loud shocking bang somewhere in the track. Get a baseline measurement initially, then measure their conductance immediately leading up to and after the bang. Apologize (say it was an unforseen technical difficulty), and then send him off and move on to the next participant.
Problems I could see with this would be that it may be tough to get by an ethics committee, and it would require a significantly higher number of participants to get the same number of trials (you would also have to control participants from communicating with each other about their experiences, otherwise people could clue into what you're doing and expectancy could creep in even here).
Anyone else have any ideas about how expectancy could be completely ruled out at the experimental design phase?  |