For each group, put all the sets a single questionnaire. assign the all the items within the each set the same BGR value so that the order of the sets is randomized. Use a skipto value on the last question within each set which takes the subject to the end of the questionnaire--thus skipping the other sets within that questionnaire. Then, when you list your four questionnaire files in the experiment file, you can randomize the order in which they occur using WGR values. Does that make sense? You might also check out unconditional skipping in the user's guide and elsewhere here on the support site.