Hi Raimo > Hello, > I wonder if anyone has tried to estimate how many subjects would be > necessary to get a significant response in a group level, and in that > case how it is calculated? you can't just use power analysis as for any other behavioural measures - there are issues in fMRI that have to be addressed - this package is quite good http://www.fmripower.org/ although based on a ROI approach. Indeed, power changes with regions of the brain and tools are being dev. (hopefully we will soon be able to use Satoru Hayasakas' toolbox see http://fmri.wfubmc.edu/cms/talkPowerSampleSizeCalculation) good luck cyril -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.