Psychometric properties of the Revised Purdue Spatial Visualization Tests: Visualization of Rotations (the Revised PSVT:R)

Working under classical test theory (CTT) and item response theory (IRT) frameworks, this study investigated psychometric properties of the Revised Purdue Spatial Visualization Tests: Visualization of Rotations (Revised PSVT:R). The original version, the PSVT:R was designed by Guay (1976) to measure spatial visualization ability in three-dimensional mental rotation. Since then, the instrument has predominantly been used in science, technology, engineering, and mathematics (STEM) education research investigating the association between the spatial ability and STEM performance of students. However, figural errors of the instrument led to its revision for this study; thus the psychometric properties of the revised version were evaluated using the data from 1022 undergraduate students across all majors at Purdue University. Within the framework of CTT, reliability evidence using Cronbach’s alpha showed item consistency in the measure, and construct validity evidence from a confirmatory factor analysis (CFA) supported the unidimensional factor structure of the construct measured by the Revised PSVT:R. Four IRT models (Rasch, 1PL, 2PL, and 3PL models) evaluated in this study, and among them, the 3PL IRT model yielded the best model-data fit. Overall, results in both testing theory frameworks indicated that item difficulty and item discrimination were within an acceptable range. However, the items arranged by the complexity in rotation were not ordered by item difficulty, suggesting the existence of other factors, such as complexity in the 3-D shapes influencing respondents’ perceived difficulty level. According to the 3PL IRT model, a certain level of guessing effect exists across all the 30 items of the Revised PSVT:R. A multiple-groups CFA (MCFA) resulted in an invariant factor structure of the items of the instrument across genders, implying no test bias against gender