Purpose: This study aims to investigate the dimensionality of the Academic Motivation Scale items by depending on the graded response model, the generalized graded unfolding model, the bifactor model and the DIMTEST. Research Methods: The Academic Motivation Scale was implemented on 1858 students who were studying at Ankara University. The fit of models was examined based on the general, person and item level model data fit statistics that were produced by the models. Findings: It was found out that the bifactor model provided the most consistent results with the theoretical foundation underlying the items. The findings revealed that the generalized graded unfolding model and the bifactor model enabled better results than the graded response model concerning to the general model data fit. About item fit statistics, the models that provided the best fit were the bifactor model, the generalized graded unfolding model and the graded response model, respectively. The index values obtained based on the bifactor model also brought out the existence of a strong general dimension on which the scale items could be ordered. The results of DIMTEST analysis also supported that the scale items are multidimensional. Implications for Research and Practice: Researchers are recommended to estimate item parameters both on the general dimension and subscales of the Academic Motivation Scale by utilizing the bifactor model to obtain more reliable and valid item parameter estimations. In future studies, researchers can compare the models about dimensionality and monotonicity assumptions based on scales developed to measure different affective traits.