You could use a simple "control" contrast. For example, "all task condition vs baseline/fixation". That should show activation in some standard areas, and you could discard the subject if it didn't.
You have to be careful, though. If you try too hard to screen subjects using the results of the first level analysis, you could bias your results. An extreme example would be discarding all subjects who don't meet your hypothesis regarding where the activations should be.
|