Yes. Using original (up to +/-20 step change in means) and demeaned motion parameters did result in exact the same beta and T maps for all the six regressors. Thanks Rick for the detailed answer! One additional question, will the slow drift (trend) of motion parameters in each run be properly take care of by the polort bases as well? The reason I ask is SPM uses separate regressors to model each run (session). If there is no real differences between the two methods, concatenating regressors across runs would be preferred because it reduces the cost of degree of freedom.