| the parameter estimation is crushing when we consider fMRI | runs that were acquired with different number of slices. | | Any suggestions for solving this problem? I'm afraid that this is just one of SPM's limitations. All the images have to have the same dimensions, and there has to be no indication that any of them have different voxel sizes, orientations etc. This is because a voxel in one image needs to correspond with the same voxel in all the other images in the series. If you are using spatially normalised data, then make sure you use the same bounding box and voxel size for all subjects. If it is a single subject analysis, then you can realign and reslice the images so they all have the same dimansions (as the first one in the series). Best regards, -John