Dear Michael, these two p-values quantify 'surprise' about - at least partly - distinct features of the data: the existence of high, focal activations with 'well-defined' spatial localisation vs the existence of broad and extensive activations. Therefore, according the form of the putative experimentally-induced signal (focal and peaky or broad, low and extensive), peak vs cluster level inferences will be more appropriate/sensitive. Slide 6 of Tom's lecture is a good illustration: http://www.fil.ion.ucl.ac.uk/spm/course/slides09/05_Random_Field_Theory.ppt Best wishes, Guillaume. Michael Froelich wrote: > Dear SPM community, > > my question relates to random field theory and inference. > > In the SPM results table, what is the relationship of the two p_FWE-corr, the family-wise > corrected p-value reported under the heading "peak-level" and the one reported under the > heading "cluster-level". > > In particular, what is the meaning of a significant cluster-level p_FWE-corr at the 0.05 level > when the corresponding peak-level p_FWE-corr is not significant at the 0.05 level? > > Michael Froelich -- Guillaume Flandin, PhD Wellcome Trust Centre for Neuroimaging University College London 12 Queen Square London WC1N 3BG