Is There an Optimum of Realism in Computer-Generated Instructional Visualizations?

Realistic visualizations are abundantly used in digital education. However, the use of realism is still thought to risk a cognitive overload due to excessive details. Moreover, it is still not precisely known whether there is an optimal level of realism that benefits learners the most. In two experimental studies, different versions of anatomical visualizations were compared regarding their effects on retention performance and the subjective cognitive load experienced during learning. In Experiment 1 (n = 73), four visualizations with minor variations in the geometry and shading of the model featured in the visualizations were used. Although neither the level of detail in the geometry nor the realism of the shading resulted in significant differences, a detailed model with simplified shading elicited the highest retention scores descriptively. In Experiment 2 (n = 156), a schematic visualization was compared with an "idealized" model featuring only simplified shading and a highly realistic rendering. There most realistic version elicited the highest retention scores, but also the highest cognitive load ratings. Taken together, the results suggest that the optimal level of realism might lie on the more realistic end of the spectrum for learning tasks focused on the memorization of shapes that are assessed using image-based tests.