Gender differences in three-dimensional mental rotation are well documented in literature. The item material itself has been shown to affect these gender differences as well as general performance in mental rotation tasks. In this thesis, a mental rotation test based on automatic item generation was tested. The item material in question were endless loops. While previous research suggested Rasch homogeneity for endless loops, further and stricter assessment of the Rasch model is still needed. For this purpose, 25 new endless loops were generated. Four endless loops were specifically designed to cause Rasch heterogeneity. Rasch heterogeneity was expected for all endless loops, while excluding the four Rasch heterogeneous endless loops should lead to Rasch homogeneity. To test this hypothesis, the 25 endless loops were administered to a sample of 222 psychology students as a power test. Parametric and non-parametric Andersen Likelihood Ratio tests and Martin-Löf tests with alpha=5% were computed as goodness-of-fit tests for a 1PL Rasch model. As expected, all endless loops together were Rasch heterogeneous. However, exclusion of the four endless loops designed to elicit Rasch heterogeneity did not suffice to achieve Rasch homogeneity; two more endless loops had to be excluded. Post-hoc analyses of the two deviating endless loops are discussed in regard to (1) the basic alignment of the endless loops and non-spatial strategies, (2) relationship of depth cue salience with gender differences being linear or non-linear and (3) action affordance.