Hi SPM list, Each of our subjects has four functional runs/tasks, each of which I preprocess separately using a copy of the same T1 structural scan. Sometimes the segmentation does not work (see BadSegment.jpg, attached), even when the segmentation was successful when coregistered to the mean of a different functional run (see GoodSegment.jpg, attached, from the same structural scan). From realignment parameters, this subject did not move more than 2mm in any direction between functional runs. This has happened for 5 of 27 subjects so far. Is there something I can do to avoid getting this bad segmentation? I am running SPM5 Updates1111, Matlab 7.4.0 (r2007b), on Linux. Thanks for any help, Kyle Shattuck CFMI, Georgetown University