Dear Jue Mo, >I've a question about the DCM application. DCM already been applied >to the steady state response, and the neuronal response only relies >on the cross spectra, why didn't you apply it to the ongoing activity? That is an interesting suggestion. We have not pursued this because the spectral density of fMRI responses will be determined largely by the parameters of the hemodynamic response function (and less by the effective connectivity among regions). Note that DCM for Steady-State responses is a model for spectral density data (not the time course). This can be very efficient when dealing with long time-series but throws away many degrees of freedom in the data. We are currently working on DCM for ongoing activity (e.g., resting state time-series) using stochastic DCM (and generalized filtering). This should be ready in about a month or so. And might be usefully compared with Ganger Causal Analyses. With very best wishes, Karl