A concept for a so-called matrices test (a missing figure has to be deduced by reasoning, according to row- and/or column-wise logically proceeding figures in a matrix) is introduced which is new in the following ways: 1) Instead of using a typical 3x3 matrix design, a 5x5 matrix design is used, 2) in contrast to common matrices, highly redundant cells are emptied, and 3) the questioned cell is located rather at random instead of in the last row and the last column. Due to the well-known problems of a multiple-choice response format (i.e. recognizing the solution instead of producing; the phenomenon of lucky guessing; construction of plausible distractors), a free response format is used: The testee has to draw the solution by hand, whereby the necessary drawing ability is extremely low. A first draft of the Two-way Figural Reasoning-Test essentially measures in accordance with the Rasch model. However, testing whether the given item generating rules sufficiently explain the resulting item difficulty parameters by means of the LLTM (linear logistic test model) disclosed that some up to now non-aware (cognitive) operation components interfere with the items’ difficulty. Above all, further research is needed in order to remove any not reasoning-based radicals of item difficulty – some suggestions of such radicals are given.