Dear SPM'ers When is it reasonable to perform a subject specific subtraction analysis (Multi-subj: cond x subj interaction & covariates) as fixed effects analysis? Is it reasonable for example in a study of emotion (2 conditions, 4 replications/cond, 88 scans)? We used a contrast [1 -1 ... 1 -1] (contrast vector of 22 components, 11 subjects) to test the average difference. Is this correct? Esa Wallius Turku PET Centre, Finland