Dear Cinzia, This sounds like a bug that was present in the SPM5 updates 1111: http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0802&L=spm&P=10382 To check which SPM version you have, type: >> edit spm_getSPM.m and around line 162, you should read: % $Id: spm_getSPM.m 1563 2008-05-07 14:16:43Z ferath $ If not, you should install the latest set of updates from: ftp://ftp.fil.ion.ucl.ac.uk/spm/spm5_updates/ Note that if that was indeed the explanation, then your SPM.mat files have been corrupted and you'd better re-estimate your models. I hope this helps, Guillaume. Cinzia Di Dio wrote: > Dear all > > I’m writing to ask for advice on diagnosing a systematic problem in > my SPM analyses. I’m running SPM5 with Matlab 2007b on a Windows XP > machine. For the past two months I have been unable to run first > level analyses because of systematic unpredictable errors. > > When I try to recall the data from a SPM.mat file (after either first > or second level processing) in some cases the recalled file does not > correspond with the original data. This error does not affect all > files, I have been unable to establish why some file mappings are > wrong while other remain correct. This problem seems completely > random and thus I am unable to pinpoint the source. > > I am also encountering a problem with some of the contrast images > (again a random selection). The images are totally degraded, highly > pixelated, no activation pattern is visible, and I can only say they > almost look to have been scribbled over (in black and white). On some > occasions this problem shows up when the contrast image is first > made. > > I am quite accustomed to running these analyses and I am sure there > is no problem with the data or the machine itself. It appears that > SPM / Matlab could be rewriting file paths, at least for the first > problem? > > I appreciate that you are unable to offer one to one SPM support, but > I am having extreme difficulties in pinpointing the source of these > errors, particularly as they are irregular. I would be very grateful > for any information you might be able to provide to unable me to > solve these problems. > > Thank you very much > > Best regards Cinzia Di Dio University of Parma, Italy -- Guillaume Flandin, PhD Wellcome Trust Centre for Neuroimaging University College London 12 Queen Square London WC1N 3BG