There have been myriad conceptual inventories developed and used by the physics education research community (e.g., FCI, FMCE, BEMA, CSEM). However, continued examination of the psychometric properties of these inventories is important for improving inventory development and usage practices. In this study, we investigate the dimensionality of the Energy and Momentum Conceptual Survey (EMCS) test items and explore the structure of students' understanding of these concepts (i.e., whether the student response data showed multiple distinct test traits). To investigate the psychometric properties of this survey, we surveyed a sample of 253 students participated in the pretest and 201 students for the posttest assessments. Statistical analyses, guided by Item Response Theory (IRT), were conducted using R software. The results of Bootstrap modified parallel analysis tests (BMPAT) revealed significant differences between unidimensional models and the actual data, indicating the presence of multidimensionality (i.e., dimensions correspond to the
physics concepts/abilities being tested) in the EMCS test items. Exploratory analyses suggested that a 2 or 3-dimensional model best fits the data. However, categorizing items based on designated concepts did not improve the model fit. The findings imply that students may interpret the items differently than intended by the test designers.