Hi Dana,

One possible reason for this would be SPM's "implicit masking" setting during realignment, where any voxels that are not inside of the field of view at all timepoints are set to NaN/0 (so if at one individual point the subject moved "a lot" and the acquisition field-of-view of that individual scan did not include the whole brain, then the entire timeseries ends up being similarly cropped). There are of course good reasons to do that (e.g. to avoid having interpolated BOLD timeseries with "missing" data at individual timepoints), but, assuming that you are using ART or other similar procedures to tag and properly deal with outlier scans, then it may be reasonable to skip this masking. In CONN's development release I have added a field to optionally disable this implicit masking (see BATCH.Setup.preprocessing.rmask option in "help conn_batch"). If you want to try that out and see whether that fixes this issues, the syntax to run realign and unwarp on your project's current functional data without implicit masking would be:

conn_batch( 'Setup.preprocessing.steps', {'functional_realign&unwarp'}, ...
                   'Setup.preprocessing.rmask', false);

Hope this helps
Alfonso

Originally posted by Dana Glenn: