Measurement of spatial ability: Reevaluating and Example of test items from the flags revising measures of mental rotation

In order to accurately assess and characterize spatial ability, commonly used measures of this ability need to be evaluated psychometrically and compared to determine whether they measure the same or different constructs (Uttal, McKee, Simms, Hegarty, & Newcombe, 2024). Here we examined computer-administered versions of three commonly used tests of mental rotation ability, the Flags Test (Thurstone & Jeffrey, 1959), the Mental Rotations Test (Vandenberg & Kuse, 1978), and the Revised Purdue Spatial Visualization Test ( Yoon, 2011 ). On the basis of Item Response Theory (IRT) analyses, and concerns about the validity of some test items, we created shortened and refined versions of these tests, which maintained the high reliability of the original tests. The three refined test versions were then combined into a 35-item hybrid mental rotation test. The hybrid test showed high reliability and general factor saturation, and was best fit by a Hierarchical Bifactor IRT model, with all items loading on a general ability factor, and the items from each test also loading on subfactors reflecting test-specific differences. No evidence for a separate ability to mentally rotate 3-dimensional vs. 2-dimensional figures was found. The hybrid test showed high precision across a wide range of ability (±2 standard deviations from average) and a small gender difference in favor of men. The hybrid test functions well as an efficient measure of mental rotation ability across a wide range of difficulty that is less confounded by strategy, visual perception, and time constraints than the original individual tests.