For running a Random Effects analysis, the separation of effects between subjects is unavoidable. When using the standard Multi-study GLM approach in BrainVoyager, this means one has to use the “separate subjects” check-mark
The drawback is that this method averages the data of the subjects over their different runs.
Using the “separate studies” check-mark on the other hand will not be the proper option to run a Random Effects Analysis because one is separating all the runs (which may be multiple for each subject). When one wants to compare the different runs in the course of the Random Effects analysis, there is an easy workaround that allows to compare predictors as well as runs in the course of statistical contrasts.
It is all described in this PDF document.