Hi there, > I have a question about the FWE analysis. > Is this correction Bonferroni correction? > Or, SPM use another type of correction? (e.g. Monte Carlo simulations) I'm not the expert on this but since I'm here .. FWE like Bonferroni relies on a max statistic, i.e. find a threshold value and apply. Whereas Bonferroni will correct for the nb of observations (nb of voxels) SPM proposes to correct for the nb of independante observations which depends on the smoothness of your image (see random field theory http://imaging.mrc-cbu.cam.ac.uk/imaging/PrinciplesRandomFields) > > And, if I use uncorrected 0.001 with 3 voxels (it's cluster-size > thresholding, right?), how is it statistically different from > corrected 0.05 without cluster-size thresholding? what do you mean with 3 voxels - in the spm table you have the cluster inference available (the extend threshold is more like a clean up than a statistical threshold) details on the different levels of inference on a 'old' article www.fil.ion.ucl.ac.uk/spm/doc/papers/karl_detecting.pdf best cyril -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.