Engagement is not encoding: Outcome alignment determines the effectiveness of mixed reality in medical education

Mixed Reality (MR) technologies are increasingly discussed as a resource-efficient option to support high-quality teaching in medical education. Yet the literature is still dominated by feasibility studies and systematic reviews; controlled comparative studies remain less common. The study aimed to assess whether mixed reality provides a learning benefit in a single, time-limited teaching session and to characterise students' perceptions and how these change from pre-to post-session. An exploratory questionnaire was developed in advance based on semi-structured, guide-based interviews (n = 15). Subsequently, 46 medical and dental students were randomised to either an intervention group (mixed reality; n = 23) or a control group (script-based conventional teaching; n = 23). After the respective learning unit (a clinically oriented anatomy session), a knowledge recall test was administered. Subjective ratings in the domains of prior knowledge, educational relevance, visualisation, understanding, interaction, technical aspects and critical feedback were collected using pre- and post-questionnaires to capture changes in perception and the extent of opinion change. In the between-group comparison, no consistent differences were observed pre- or post-intervention, nor with regard to the extent of opinion change. Within-group pre-post analyses assessed perception change. In the control group, significant pre-post changes were found in 6 of 7 domains (prior knowledge p < 0.001; educational relevance p < 0.001; visualisation p < 0.001; understanding p < 0.001; interaction p = 0.023; technical aspects p = 0.004). In contrast, in the critical feedback domain, a significant difference was observed only in the intervention group (p = 0.029). In the knowledge recall test, the control group performed significantly better (p = 0.008). These findings suggest that immersive technologies do not automatically enhance learning performance. Rather, their effectiveness appears to depend on the alignment between medium characteristics, cognitive demands, and the specific learning outcomes assessed. Engagement and interactivity alone may not translate into improved short-term recall, particularly when outcomes primarily measure factual retrieval.