Three-dimensional (3D) digital anatomical models show potential to demonstrate complex anatomical relationships; however, the literature is inconsistent as to whether they are effective in improving the anatomy performance, particularly for students with low spatial visualization ability (Vz). This study investigated the educational effectiveness of a 3D stereoscopic model of the pelvis, and the relationship between learning with 3D models and Vz. It was hypothesized that participants learning with a 3D pelvis model would outperform participants learning with a two-dimensional (2D) visualization or cadaveric specimen on a spatial anatomy test, particularly when comparing those with low Vz. Participants (n = 64) were stratified into three experimental groups, who each attended a learning session with either a 3D stereoscopic model (n = 21), 2D visualization (n = 21), or cadaveric specimen (n = 22) of the pelvis. Medical and pre-medical student participants completed a multiple-choice pre-test and post-test during their respective learning session, and a long-term retention (LTR) test 2 months later. Results showed no difference in anatomy test improvement or LTR performance between the experimental groups. A simple linear regression analysis showed that within the 3D group, participants with high Vz tended to retain more than those with low Vz on the LTR test (R[superscript 2] = 0.31, P = 0.01). The low Vz participants may be cognitively overloaded by the complex spatial cues from the 3D stereoscopic model. Results of this study should inform resource selection and curriculum design for health professional students, with attention to the impact of Vz on learning.