Learning with augmented reality: Impact of dimensionality and spatial abilities

•Dimensionality of AR-based visualizations in education influences learning.•The study compares learning with a 3D or a 2D visualization of a human heart in AR.•Germane cognitive load and spatial learning outcomes are increased for the 3D group.•Learners with high but not low mental rotation abilities profited from 3D model.•High spatial abilities may provide the necessary skills to learn with a 3D AR model. Three-dimensional (3D) representations are often more effective for learning about spatial objects than two-dimensional (2D) representations. In augmented reality (AR), which can include 2D and 3D visualizations, learners’ perceptions might differ from other visual media. To examine the specific influence of the dimensionality of AR visualizations on learning the spatial structure of components, 3D and 2D AR representations of the human heart were compared in an experimental laboratory study, otherwise keeping the conditions as comparable as possible. The participants (N = 150) received the respective AR representation and were instructed to look for the hearts’ components mentioned in an informational text. As expected, learning with the 3D compared to the 2D representation resulted in higher germane cognitive load and knowledge about spatial relations of components. Proposed effects on extraneous cognitive load, knowledge about spatial positions of components and mediation effects were not found. Higher mental rotation abilities were found to be more beneficial for learning with the 3D visualization, suggesting that these learners were better equipped for this task and supporting an ability-as-enhancer hypothesis. Overall, the study revealed that even in AR scenarios, 3D visualizations may be better to convey knowledge about spatial structures than 2D visualizations. Moreover, the results emphasize the specific moderating role of spatial abilities when learning with 3D AR material. In future research with various spatial learning material and spatial abilities, the generalizability of these results needs to be examined, so that ultimately more insights can be gained for the design of optimal AR learning experiences.