Hi Anderson, Thanks for your answer. If I understand correctly, I have to set up the contrast (S11 vs S21) vs (S12 vs S22) in the lower-level analysis as well? And is it necessary to include masking for this contrast? Best, Esther From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Anderson M. Winkler Sent: vrijdag 28 augustus 2015 9:17 To: [log in to unmask] Subject: Re: [FSL] Three-level-analysis FSL Hi Esther, In the 1st level, the contrasts you already have are: S11 vs S21: 1 0 -1 0 S12 vs S22: 0 1 0 -1 Now you'd just add, for each subject: (S11 vs S21) vs (S12 vs S22): 1 -1 -1 1 It should do what you need. These can be passed up to the higher level and used to make group inferences. All the best, Anderson On 27 August 2015 at 14:23, Esther Heckendorf <[log in to unmask]<mailto:[log in to unmask]>> wrote: Hi, I have a question about the set up of a three level analysis in FSL. I have two stimuli with each two levels (S11, S12, S21, S22) and I have created the contrasts S11 vs. S21 and S12 vs. S22 in the lower-level analyses. Now, I would like to calculate the following contrast: (S11 vs. S21) vs. (S12 vs. S22). Is it possible to calculate this contrast in a higher-level analysis using the COPEs of the contrasts S11 vs. S21 (contrast 1) and S12 vs. S22 (contrast 2) as inputs? For an example with 3 participants I would set up the following model: EV1 EV2 EV3 Input 1 1 0 0 Input 2 0 1 0 Input 3 0 0 1 Input 4 -1 0 0 Input 5 0 -1 0 Input 6 0 0 -1 Input 1 = First contrast for subject 1 Input 2 = First contrast for subject 2 Input 3 = First contrast for subject 3 Input 4 = Second contrast for subject 1 Input 5 = Second contrast for subject 2 Input 6 = Second contrast for subject 3 Contrasts and F-tests: EV1 EV2 EV3 subject 1 1 0 0 subject 2 0 1 0 subject 3 0 0 1 Is it possible to calculate the contrast (S11 vs. S21) vs. (S12 vs. S22) in this way or is it necessary to set up the model for each subject separately? Thanks, Esther