Hi Cesar,
> Could 3dMEMA be the way to go even though the inputs don't have information about within-subject variability (i.e. they are correlation coefficient)?
No, don't use 3dMEMA in this case since it would not work without the reliability information.
> Is there any demo script (or message board) for tests with 3 groups?
3dANOVA is probably the most straightforward approach.
Gang