SPM applies a signal intensity mask to the data to avoid running statistics on stuff outside the brain or where there is poor signal. This is likely why the masks are 'shrinking' in some cases. In the second level analysis, only voxels in which ALL participants have data are included. There are two issues here. Firstly, is there a normalization problem in some participants? You should manually check the preprocessed EPI data, especially for those such as Sub2 below who have lots of data loss. Secondly, you can try reducing the threshold at which SPM applies this masking. This is the masking thresholding applied in the first level. Default is 0.8, I tend to use a lower value (e.g. 0.5). BUT: This is a controversial thing to do which others on the list will (not unreasonably) advise against. The best thing is to find the root cause in the data for the signal reductions around the edge. Especially to confirm if the normalizations look good. Good Luck Colin Hawco, PhD Neuranalysis Consulting Neuroimaging analysis and consultation <http://www.neuranalysis.com> www.neuranalysis.com [log in to unmask] From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On Behalf Of yang ping Sent: April-04-16 2:18 AM To: [log in to unmask] Subject: [SPM] Coregister or Normalise problem Dear SPM experts, I find the contrast map (p value=1) from second level is smaller than the template. So, why the map does not fill the template? [SECOND LEVEL MAP] SECOND LEVEL MAP.jpg I wonder this issue may be came from each subject's first level map. Here is two subjects' first level map. The map of SUB1 may looks good, but the map of SUB2 looks smaller than the template. So what can I do? Can I trust the second level results? [SUB1, FIRST LEVEL] SUB1_FIRSTLEVEL.jpg [SUB2, FIRST LEVEL] SUB2_FIRSTLEVEL.jpg