Dear Fred,
This reviewer seems to be referring to the problem of artificial anti-correlations in the context of global signal regression for resting state fMRI data (i.e., resting state data are sometimes orthogonalized to the global mean brain signal using the global signal as a confound regressor in a GLM, prior to the actual functional connectivity analyses). For details, see this paper by Murphy et al.:
http://www.ncbi.nlm.nih.gov/pubmed/18976716This is not an issue for DCM. DCM for fMRI is only informed about local signals (i.e., those from your regions of interest) and does not consider anything related to global brain signal (unless you manually construct an input that
reflects global brain signal).
Best wishes
Klaas