Hi Gang,
I have a Power Mac with 2 2.8 GHz G5 processors running Mac OS 10.4.11. I'll the multiple CPU option on my home Linux box this evening to see if I can duplicate the error. Would it have anything to do with the version of R?
In terms of the results compared to 3dttest there generally similar in that the same regions are activated but 3dMEMA yields considerably more power. If you look at the attached jpeg, it is the subtraction of the t statistic for the 3dMEMA comparison minus the t statistic 3dttest contrast (red >= 2.8, 1.7 <= orange < 2.8; 0.5 <= yellow < 1.7; -0.5 <= green < 0.5). It seems like 3dMEMA generally produces a larger t statistic (on average .5 to 1 unit greater; see attache histogram).
Vinny