I'm finding that for some runs in some subjects, after I normalize relative to the voxel mean, as described in howto # 5, I get wildly varying signal changes in my most inferior slice. Whereas voxels elsewhere in my datasets typically have values within a percentage point or two of the voxel mean (100), the signal in voxels contained in the most inferior slice are all over the place (e.g., a minimum of 79 and a peak of 594 in one voxel for one subject). How typical is this? Is this just some sort of motion artifact? Should I just discard this slice in all subjects? I can provide data and the scripts used for normalization if this isn't something others have encountered.

Thanks!

Phil