Advances in computational technology provide opportunities to explore new methods to improve spatial abilities and the understanding of buildings in the domain of architecture education. This research devised and tested a learning approach that utilizes Augmented Reality (AR), Building Information Modeling (BIM), and physical buildings for architecture education. The research employed a quasi-experimental research design as a research method and developed a BIM-enabled AR educational tool (BIMxAR) with three prototypes: BIMxAR 1.0 to support the learning and understanding of building construction systems, materials configuration, and 3D section views of complex building structures; BIMxAR 2.0 to support learning specific building construction systems using the BIM visualization mode "visibility of categories"; and BIMxAR 3.0 to support learning layers or elements of building components and their assembly. Moreover, the research developed a novel and accurate AR registration method (DL-3S-BIM) for the scale of buildings and a novel visualization mode, which the research claims that it may improve users' spatial awareness. Furthermore, for the purpose of a comparative study, the research developed a non-AR version of each prototype, where each version is identical to the AR version in terms of the BIM example project and capabilities except missing the AR registration function. The research approach was validated through a test case, in which BIMxAR 1.0 was used as an intervention. Two study groups were employed - non-AR and AR. The recruited participants were students from the Department of Architecture and Department of Construction Science in the College of Architecture at Texas A&M University. The test case utilizes a longitudinal study approach as a data collection strategy, i.e., pretest phase-learning phase posttest phase. The research analyzed qualitative and quantitative data collected from tests, questionnaires, surveys, and observations. Considering the short learning session compared to the examples from the literature, the findings of the study show that the intervention, the AR version, contributed to higher learning gain; however, the learning gain differences between the groups were not statistically significant. The overall workload associated with the AR version was significantly lower in the AR group. Moreover, the AR version was rated significantly higher in terms of performance. Observations in the learning sessions showed that students in the non-AR group tried to align the virtual BIM model with the physical building, which was difficult, while the students in the AR group conveniently utilized the automatic alignment between the virtual BIM model and the physical building, enabled by the developed AR registration method. These findings suggest that the AR version is an easy, useful, and convenient learning tool.
