In Lillehammer we reported the results of an experiment using dynamic shading to improve completion of a grid or matrix question. In this paper we report on additional experiments on the design of such questions. We compare a standard grid with 13 items and 2 questions per item to two versions using dynamic shading and two versions in which the grid is split into separate parts, one for each of the questions. The survey was administered to members of an opt-in panel in the U.S. We report the results of this experiment and discuss the implications for design of grid or matrix questions.