> Thanks very much for your response. I believe the original ROIs were based on earlier work - which seems good. My interest is with the apparently arbitrary values selected subsequently. If you have an a priori ROI and find no robust effect when you constrain your analysis to that ROI, is it valid to then 'search' around that ROI, find a weak effect (p 0.05 uncorrected) and then use that to define your new ROI?
you are right that regardless of how the locations were originally defined the search for a local maxima means that the methods does not yield a kosher SVC. It wouldn't be easy to work out what the true false positive rate would be in a case like that.