Thank you very much Suzanne for the detailed report. This is indeed weird, as ART should not impact that much the analyses (it only adds 1st-level covariates for motion correction, the 2nd-level covariates are only for QA and should have no impact at all).

For those having a similar issue, it would be interesting to try to do the analysis after removing the 1st-level covariates generated by ART, normally this should fix the issue if ART is the culprit (but of course the motion correction would be lost but this would help in confirming the culprit).

Best regards,
Stephen