Dear Benjamin,
SPM seems to be unable to estimate useful error covariances for your model and your data. There are many possible causes for this, the most likely ones including problems with the data or using an unsuitable model. You should check the data for that subject using e.g. ArtRepair or TSDiffAna to find out whether there are 3D images or single slices that show suspicious signal. Also, running spm_imcalc on the timeseries to produce mean or std images might help. In addition, you should check your model to make sure it fits the expected signal timecourse.
Hope this helps
Volkmar
|