Hello Kirsten,
as Donald already said: It might well be that you have quite consistent effects for the individual first-level analyses. In case the number of subjects is low (9 is low) you might not have enough power to detect these effects in a random-effects second-level analysis.
Concerning a fixed-effects first-level analysis based on all your subjects, a single subject with extremely large effects might be sufficient to get significant results on group level.
Now, in your case, what happens when lowering the voxel threshold in your second-level analysis e.g. to .005 (uncorrected)? Does the expected activation reappear?
Best,
Gabor
|